2023
Journal article
Open Access
Efficient adaptive ensembling for image classification
Bruno A., Moroni D., Martinelli M.
In recent times, except for sporadic cases, the trend in Computer Vision is to achieve minor improvements over considerable increases in complexity.To reverse this tendency, we propose a novel method to boost image classification performances without an increase in complexity.To this end, we revisited ensembling, a powerful approach, not often adequately used due to its nature of increased complexity and training time, making it viable by specific design choices. First, we trained end-to-end two EfficientNet-b0 models (known to be the architecture with the best overall accuracy/complexity trade-off in image classification) on disjoint subsets of data (i.e. bagging). Then, we made an efficient adaptive ensemble by performing fine-tuning of a trainable combination layer. In this way, we were able to outperform the state-of-the-art by an average of 0.5\% on the accuracy with restrained complexity both in terms of number of parameters (by 5-60 times), and FLoating point Operations Per Second (by 10-100 times) on several major benchmark datasets, fully embracing the green AI.Source: EXPERT SYSTEMS (ONLINE)
DOI: 10.1111/exsy.13424
Metrics:
Bruno A., Moroni D., Martinelli M.
In recent times, except for sporadic cases, the trend in Computer Vision is to achieve minor improvements over considerable increases in complexity.To reverse this tendency, we propose a novel method to boost image classification performances without an increase in complexity.To this end, we revisited ensembling, a powerful approach, not often adequately used due to its nature of increased complexity and training time, making it viable by specific design choices. First, we trained end-to-end two EfficientNet-b0 models (known to be the architecture with the best overall accuracy/complexity trade-off in image classification) on disjoint subsets of data (i.e. bagging). Then, we made an efficient adaptive ensemble by performing fine-tuning of a trainable combination layer. In this way, we were able to outperform the state-of-the-art by an average of 0.5\% on the accuracy with restrained complexity both in terms of number of parameters (by 5-60 times), and FLoating point Operations Per Second (by 10-100 times) on several major benchmark datasets, fully embracing the green AI.Source: EXPERT SYSTEMS (ONLINE)
DOI: 10.1111/exsy.13424
Metrics:
See at:
CNR IRIS 
| onlinelibrary.wiley.com | ISTI Repository | CNR IRIS 
| CNR IRIS
2023
Conference article
Open Access
Revisiting ensembling for improving the performance of deep learning models
Bruno A, Moroni D, Martinelli M
Ensembling is a very well-known strategy consisting in fusing several different models to achieve a new model for classification or regression tasks. Over the years, ensembling has been proven to provide superior performance in various contexts related to pattern recognition and artificial intelligence. Moreover, the basic ideas that are at the basis of ensembling have been a source of inspiration for the design of the most recent deep learning architectures. Indeed, a close analysis of those architectures shows that some connections among layers and groups of layers achieve effects similar to those obtainable by bagging, boosting and stacking, which are the well-known three basic approaches to ensembling. However, we argue that research has not fully leveraged the potential offered by ensembling. Indeed, this paper investigates some possible approaches to the combination of weak learners, or sub-components of weak learners, in the context of bagging. Based on previous results obtained in specific domains, we extend the approach to a reference dataset obtaining encouraging results.DOI: 10.1007/978-3-031-37742-6_34
Metrics:
Bruno A, Moroni D, Martinelli M
Ensembling is a very well-known strategy consisting in fusing several different models to achieve a new model for classification or regression tasks. Over the years, ensembling has been proven to provide superior performance in various contexts related to pattern recognition and artificial intelligence. Moreover, the basic ideas that are at the basis of ensembling have been a source of inspiration for the design of the most recent deep learning architectures. Indeed, a close analysis of those architectures shows that some connections among layers and groups of layers achieve effects similar to those obtainable by bagging, boosting and stacking, which are the well-known three basic approaches to ensembling. However, we argue that research has not fully leveraged the potential offered by ensembling. Indeed, this paper investigates some possible approaches to the combination of weak learners, or sub-components of weak learners, in the context of bagging. Based on previous results obtained in specific domains, we extend the approach to a reference dataset obtaining encouraging results.DOI: 10.1007/978-3-031-37742-6_34
Metrics:
See at:
CNR IRIS | link.springer.com | ISTI Repository | CNR IRIS | CNR IRIS
2023
Contribution to book
Open Access
Mesoscale events classification in sea surface temperature imagery
Reggiannini M, Janeiro J, Martins F, Papini O, Pieri G
Sea observation through remote sensing technologies plays an essential role in understanding the health status of marine fauna species and their future behaviour. Accurate knowledge of the marine habitat and the factors affecting faunal variations allows to perform predictions and adopt proper decisions. This is even more relevant nowadays, with policymakers needing increased environmental awareness, aiming to implement sustainable policies. There is a connection between the biogeochemical and physical processes taking place within a biological system and the variations observed in its faunal populations. Mesoscale phenomena, such as upwelling, countercurrents and filaments, are essential processes to analyse because their arousal entails, among other things, variations in the density of nutrient substances, in turn affecting the biological parameters of the habitat. This paper concerns the proposal of a classification system devoted to recognising marine mesoscale events. These phenomena are studied and monitored by analysing Sea Surface Temperature images captured by satellite missions, such as Metop and MODIS Terra/Aqua. Classification of such images is pursued through dedicated algorithms that extract temporal and spatial features from the data and apply a set of rules to the extracted features, in order to discriminate between different observed scenarios. The results presented in this work have been obtained by applying the proposed approach to images captured over the south-western region of the Iberian Peninsula.DOI: 10.1007/978-3-031-25599-1_38
Project(s): NAUTILOS
Metrics:
Reggiannini M, Janeiro J, Martins F, Papini O, Pieri G
Sea observation through remote sensing technologies plays an essential role in understanding the health status of marine fauna species and their future behaviour. Accurate knowledge of the marine habitat and the factors affecting faunal variations allows to perform predictions and adopt proper decisions. This is even more relevant nowadays, with policymakers needing increased environmental awareness, aiming to implement sustainable policies. There is a connection between the biogeochemical and physical processes taking place within a biological system and the variations observed in its faunal populations. Mesoscale phenomena, such as upwelling, countercurrents and filaments, are essential processes to analyse because their arousal entails, among other things, variations in the density of nutrient substances, in turn affecting the biological parameters of the habitat. This paper concerns the proposal of a classification system devoted to recognising marine mesoscale events. These phenomena are studied and monitored by analysing Sea Surface Temperature images captured by satellite missions, such as Metop and MODIS Terra/Aqua. Classification of such images is pursued through dedicated algorithms that extract temporal and spatial features from the data and apply a set of rules to the extracted features, in order to discriminate between different observed scenarios. The results presented in this work have been obtained by applying the proposed approach to images captured over the south-western region of the Iberian Peninsula.DOI: 10.1007/978-3-031-25599-1_38
Project(s): NAUTILOS
Metrics:
See at:
CNR IRIS | link.springer.com | ISTI Repository | CNR IRIS | CNR IRIS
2023
Conference article
Open Access
On the applicability of prototypical part learning in medical images: breast masses classification using ProtoPNet
Carloni G, Berti A, Iacconi C, Pascali Ma, Colantonio S
Deep learning models have become state-of-the-art in many areas, ranging from computer vision to agriculture research. However, concerns have been raised with respect to the transparency of their decisions, especially in the image domain. In this regard, Explainable Artificial Intelligence has been gaining popularity in recent years. The ProtoPNet model, which breaks down an image into prototypes and uses evidence gathered from the prototypes to classify an image, represents an appealing approach. Still, questions regarding its effectiveness arise when the application domain changes from real-world natural images to gray-scale medical images. This work explores the applicability of prototypical part learning in medical imaging by experimenting with ProtoPNet on a breast masses classification task. The two considered aspects were the classification capabilities and the validity of explanations. We looked for the optimal model's hyperparameter configuration via a random search. We trained the model in a five-fold CV supervised framework, with mammogram images cropped around the lesions and ground-truth labels of benign/malignant masses. Then, we compared the performance metrics of ProtoPNet to that of the corresponding base architecture, which was ResNet18, trained under the same framework. In addition, an experienced radiologist provided a clinical viewpoint on the quality of the learned prototypes, the patch activations, and the global explanations. We achieved a Recall of 0.769 and an area under the receiver operating characteristic curve of 0.719 in our experiments. Even though our findings are non-optimal for entering the clinical practice yet, the radiologist found ProtoPNet's explanations very intuitive, reporting a high level of satisfaction. Therefore, we believe that prototypical part learning offers a reasonable and promising trade-off between classification performance and the quality of the related explanation.DOI: 10.1007/978-3-031-37660-3_38
Metrics:
Carloni G, Berti A, Iacconi C, Pascali Ma, Colantonio S
Deep learning models have become state-of-the-art in many areas, ranging from computer vision to agriculture research. However, concerns have been raised with respect to the transparency of their decisions, especially in the image domain. In this regard, Explainable Artificial Intelligence has been gaining popularity in recent years. The ProtoPNet model, which breaks down an image into prototypes and uses evidence gathered from the prototypes to classify an image, represents an appealing approach. Still, questions regarding its effectiveness arise when the application domain changes from real-world natural images to gray-scale medical images. This work explores the applicability of prototypical part learning in medical imaging by experimenting with ProtoPNet on a breast masses classification task. The two considered aspects were the classification capabilities and the validity of explanations. We looked for the optimal model's hyperparameter configuration via a random search. We trained the model in a five-fold CV supervised framework, with mammogram images cropped around the lesions and ground-truth labels of benign/malignant masses. Then, we compared the performance metrics of ProtoPNet to that of the corresponding base architecture, which was ResNet18, trained under the same framework. In addition, an experienced radiologist provided a clinical viewpoint on the quality of the learned prototypes, the patch activations, and the global explanations. We achieved a Recall of 0.769 and an area under the receiver operating characteristic curve of 0.719 in our experiments. Even though our findings are non-optimal for entering the clinical practice yet, the radiologist found ProtoPNet's explanations very intuitive, reporting a high level of satisfaction. Therefore, we believe that prototypical part learning offers a reasonable and promising trade-off between classification performance and the quality of the related explanation.DOI: 10.1007/978-3-031-37660-3_38
Metrics:
See at:
CNR IRIS | link.springer.com | ISTI Repository | CNR IRIS | CNR IRIS
2023
Conference article
Open Access
Automated image processing for remote sensing data classification
Reggiannini M, Papini O, Pieri G
Remote sensing technologies allow for continuous and valuable monitoring of the Earth's various environments. In particular, coastal and ocean monitoring presents an intrinsic complexity that makes such monitoring the main source of information available. Oceans, being the largest but least observed habitat, have many different factors affecting theirs faunal variations. Enhancing the capabilities to monitor and understand the changes occurring allows us to perform predictions and adopt proper decisions. This paper proposes an automated classification tool to recognise specific marine mesoscale events. Typically, human experts monitor and analyse these events visually through remote sensing imagery, specifically addressing Sea Surface Temperature data. The extended availability of this kind of remote sensing data transforms this activity into a time-consuming and subjective interpretation of the information. For this reason, there is an increased need for automated or at least semi-automated tools to perform this task. The results presented in this work have been obtained by applying the proposed approach to images captured over the southwestern region of the Iberian Peninsula.DOI: 10.1007/978-3-031-37742-6_43
Project(s): NAUTILOS
Metrics:
Reggiannini M, Papini O, Pieri G
Remote sensing technologies allow for continuous and valuable monitoring of the Earth's various environments. In particular, coastal and ocean monitoring presents an intrinsic complexity that makes such monitoring the main source of information available. Oceans, being the largest but least observed habitat, have many different factors affecting theirs faunal variations. Enhancing the capabilities to monitor and understand the changes occurring allows us to perform predictions and adopt proper decisions. This paper proposes an automated classification tool to recognise specific marine mesoscale events. Typically, human experts monitor and analyse these events visually through remote sensing imagery, specifically addressing Sea Surface Temperature data. The extended availability of this kind of remote sensing data transforms this activity into a time-consuming and subjective interpretation of the information. For this reason, there is an increased need for automated or at least semi-automated tools to perform this task. The results presented in this work have been obtained by applying the proposed approach to images captured over the southwestern region of the Iberian Peninsula.DOI: 10.1007/978-3-031-37742-6_43
Project(s): NAUTILOS
Metrics:
See at:
CNR IRIS | link.springer.com | ISTI Repository | CNR IRIS | CNR IRIS
2023
Conference article
Open Access
Haptic-based cognitive mapping to support shopping malls exploration
Paratore Mt, Leporini B
This paper describes a study, which is currently underway, whose aim is to investigate how the haptic channel can be effectively exploited by visually impaired users in a mobile app for the preliminary exploration of an indoor environment, namely a shopping mall. Our goal was to use haptics to convey knowledge of how the points of interest (POIs) are distributed within the physical space, and at the same time provide information about the function of each POI, so that users can get a perception of how functional areas are distributed in the environment "at a glance". Shopping malls are typical indoor environments in which orientation aids are highly appreciated by customers, and many different functional areas persist. We identified seven typical categories of POIs which can be encountered in a mall, and then associated a different vibration pattern each. In order to validate our approach, we designed and developed a prototype for preliminary testing, based on the Android platform. The prototype was periodically debugged with the aid of two visually impaired experienced users, who gave us precious advice throughout the development process. We will describe how this app was conceived, the issues emerged during its development and the positive outcomes produced by a very early testing stage. Finally, we will show that the proposed approach is promising and is worthy of further investigation.Source: LECTURE NOTES OF THE INSTITUTE FOR COMPUTER SCIENCES, SOCIAL INFORMATICS AND TELECOMMUNICATIONS ENGINEERING (INTERNET), pp. 54-62. Online event, 16-18/11/2022
DOI: 10.1007/978-3-031-28813-5_4
Metrics:
Paratore Mt, Leporini B
This paper describes a study, which is currently underway, whose aim is to investigate how the haptic channel can be effectively exploited by visually impaired users in a mobile app for the preliminary exploration of an indoor environment, namely a shopping mall. Our goal was to use haptics to convey knowledge of how the points of interest (POIs) are distributed within the physical space, and at the same time provide information about the function of each POI, so that users can get a perception of how functional areas are distributed in the environment "at a glance". Shopping malls are typical indoor environments in which orientation aids are highly appreciated by customers, and many different functional areas persist. We identified seven typical categories of POIs which can be encountered in a mall, and then associated a different vibration pattern each. In order to validate our approach, we designed and developed a prototype for preliminary testing, based on the Android platform. The prototype was periodically debugged with the aid of two visually impaired experienced users, who gave us precious advice throughout the development process. We will describe how this app was conceived, the issues emerged during its development and the positive outcomes produced by a very early testing stage. Finally, we will show that the proposed approach is promising and is worthy of further investigation.Source: LECTURE NOTES OF THE INSTITUTE FOR COMPUTER SCIENCES, SOCIAL INFORMATICS AND TELECOMMUNICATIONS ENGINEERING (INTERNET), pp. 54-62. Online event, 16-18/11/2022
DOI: 10.1007/978-3-031-28813-5_4
Metrics:
See at:
CNR IRIS | link.springer.com | ISTI Repository | CNR IRIS | CNR IRIS
2023
Journal article
Restricted
Shallow portion of an active geothermal system revealed by multidisciplinary studies: the case of Le Biancane (Larderello, Italy)
Granieri D, Mazzarini F, Cerminara M, Calusi B, Scozzari A, Menichini M, Lelli M
The natural park of Le Biancane is located in the southern sector of the Larderello-Travale geothermal field (LTGF). It extends over an approximately 100,000 m2 area where the impermeable caprock is locally absent and deep fluids may directly reach the surface. Through a multidisciplinary approach including measurements of soil CO2 flux (total output of 11.5 t day-1), soil temperature (average 34.4 °C), stable isotope and chemical data on fluids from fumaroles (dominated by a mixture of geothermal gases and air or gases from air-saturated meteoric water), and structural analysis of the formation outcropping, we found that anomalous CO2 emissions are positively correlated with shallow temperature anomalies. These are in restricted locations adjacent to vents and fumaroles, where a network of well-connected fractures (preferentially NW-SE and NE-SW orientated and with steep dips) drains efficiently allowing upward migration of the deep fluids and the energy toward the surface.Source: GEOTHERMICS, vol. 108
DOI: 10.1016/j.geothermics.2022.102616
Metrics:
Granieri D, Mazzarini F, Cerminara M, Calusi B, Scozzari A, Menichini M, Lelli M
The natural park of Le Biancane is located in the southern sector of the Larderello-Travale geothermal field (LTGF). It extends over an approximately 100,000 m2 area where the impermeable caprock is locally absent and deep fluids may directly reach the surface. Through a multidisciplinary approach including measurements of soil CO2 flux (total output of 11.5 t day-1), soil temperature (average 34.4 °C), stable isotope and chemical data on fluids from fumaroles (dominated by a mixture of geothermal gases and air or gases from air-saturated meteoric water), and structural analysis of the formation outcropping, we found that anomalous CO2 emissions are positively correlated with shallow temperature anomalies. These are in restricted locations adjacent to vents and fumaroles, where a network of well-connected fractures (preferentially NW-SE and NE-SW orientated and with steep dips) drains efficiently allowing upward migration of the deep fluids and the energy toward the surface.Source: GEOTHERMICS, vol. 108
DOI: 10.1016/j.geothermics.2022.102616
Metrics:
See at:
CNR IRIS | CNR IRIS | CNR IRIS | www.sciencedirect.com
2023
Journal article
Open Access
Brain metastases from NSCLC treated with stereotactic radiotherapy: prediction mismatch between two different radiomic platforms
Carloni G, Garibaldi C, Marvaso G, Volpe S, Zaffaroni M, Pepa M, Isaksson Lj, Colombo F, Durante S, Lo Presti G, Raimondi S, Spaggiari Lj, De Marinis F, Piperno G, Vigorito S, Gandini S, Cremonesi M, Positano V, Jereczekfossa Ba
Background and purpose. Radiomics enables the mining of quantitative features from medical images. The influence of the radiomic feature extraction software on the final performance of models is still a poorly understood topic. This study aimed to investigate the ability of radiomic features extracted by two different radiomic platforms to predict clinical outcomes in patients treated with radiosurgery for brain metastases from non-small cell lung cancer. We developed models integrating pre-treatment magnetic resonance imaging (MRI)-derived radiomic features and clinical data. Materials and methods. Pre-radiotherapy gadolinium enhanced axial T1-weighted MRI scans were used. MRI images were re-sampled, intensity-shifted, and histogram-matched before radiomic extraction by means of two different platforms (PyRadiomics and SOPHiA Radiomics). We adopted LASSO Cox regression models for multivariable analyses by creating radiomic, clinical, and combined models using three survival clinical endpoints (local control, distant progression, and overall survival). The statistical analysis was repeated 50 times with different random seeds and the median concordance index was used as performance metric of the models. Results. We analysed 276 metastases from 148 patients. The use of the two platforms resulted in differences in both the quality and the number of extractable features. That led to mismatches in terms of end-to-end performance, statistical significance of radiomic scores, and clinical covariates found significant in combined models. Conclusion. This study shed new light on how extracting radiomic features from the same images using two different platforms could yield several discrepancies. That may lead to acute consequences on drawing conclusions, comparing results across the literature, and translating radiomics into clinical practice.Source: RADIOTHERAPY AND ONCOLOGY, vol. 178
DOI: 10.1016/j.radonc.2022.11.013
Metrics:
Carloni G, Garibaldi C, Marvaso G, Volpe S, Zaffaroni M, Pepa M, Isaksson Lj, Colombo F, Durante S, Lo Presti G, Raimondi S, Spaggiari Lj, De Marinis F, Piperno G, Vigorito S, Gandini S, Cremonesi M, Positano V, Jereczekfossa Ba
Background and purpose. Radiomics enables the mining of quantitative features from medical images. The influence of the radiomic feature extraction software on the final performance of models is still a poorly understood topic. This study aimed to investigate the ability of radiomic features extracted by two different radiomic platforms to predict clinical outcomes in patients treated with radiosurgery for brain metastases from non-small cell lung cancer. We developed models integrating pre-treatment magnetic resonance imaging (MRI)-derived radiomic features and clinical data. Materials and methods. Pre-radiotherapy gadolinium enhanced axial T1-weighted MRI scans were used. MRI images were re-sampled, intensity-shifted, and histogram-matched before radiomic extraction by means of two different platforms (PyRadiomics and SOPHiA Radiomics). We adopted LASSO Cox regression models for multivariable analyses by creating radiomic, clinical, and combined models using three survival clinical endpoints (local control, distant progression, and overall survival). The statistical analysis was repeated 50 times with different random seeds and the median concordance index was used as performance metric of the models. Results. We analysed 276 metastases from 148 patients. The use of the two platforms resulted in differences in both the quality and the number of extractable features. That led to mismatches in terms of end-to-end performance, statistical significance of radiomic scores, and clinical covariates found significant in combined models. Conclusion. This study shed new light on how extracting radiomic features from the same images using two different platforms could yield several discrepancies. That may lead to acute consequences on drawing conclusions, comparing results across the literature, and translating radiomics into clinical practice.Source: RADIOTHERAPY AND ONCOLOGY, vol. 178
DOI: 10.1016/j.radonc.2022.11.013
Metrics:
See at:
CNR IRIS | ISTI Repository | www.sciencedirect.com | CNR IRIS | CNR IRIS
2023
Journal article
Open Access
Privacy by design in systems for assisted living, personalized care and well-being: a stakeholder analysis
Carboni A, Russo D, Moroni D, Barsocchi P
The concept of privacy-by-design within a system for assisted living, personalized care and well-being is crucial to protect users from misuse of the data collected about their health. Especially if the information is collected through audio-video devices, the question is even more delicate due to the nature of this data. In fact, in addition to guaranteeing a high level of privacy, it is necessary to reassure end-users about the correct use of these streams. The evolution of data analysis techniques began to take In review on an important role and increasingly defined characteristics in recent years. In this article, with reference to European projects in the AHA/AAL domain, we will see a differentiation of the concept of privacy-by-design according to different dimensions (Technical, Contextual, Business) and to the Stakeholders involved. The analysis is intended to cover technical aspects, legislative and policies-related aspects also regarding the point of view of the municipalities and aspects related to the acceptance and, therefore, to the perception of the safety of these technologies by the final end-users.Source: FRONTIERS IN DIGITAL HEALTH
DOI: 10.3389/fdgth.2022.934609
Project(s): PlatformUptake.eu
Metrics:
Carboni A, Russo D, Moroni D, Barsocchi P
The concept of privacy-by-design within a system for assisted living, personalized care and well-being is crucial to protect users from misuse of the data collected about their health. Especially if the information is collected through audio-video devices, the question is even more delicate due to the nature of this data. In fact, in addition to guaranteeing a high level of privacy, it is necessary to reassure end-users about the correct use of these streams. The evolution of data analysis techniques began to take In review on an important role and increasingly defined characteristics in recent years. In this article, with reference to European projects in the AHA/AAL domain, we will see a differentiation of the concept of privacy-by-design according to different dimensions (Technical, Contextual, Business) and to the Stakeholders involved. The analysis is intended to cover technical aspects, legislative and policies-related aspects also regarding the point of view of the municipalities and aspects related to the acceptance and, therefore, to the perception of the safety of these technologies by the final end-users.Source: FRONTIERS IN DIGITAL HEALTH
DOI: 10.3389/fdgth.2022.934609
Project(s): PlatformUptake.eu
Metrics:
See at:
CNR IRIS | ISTI Repository | ISTI Repository | www.frontiersin.org | CNR IRIS | CNR IRIS
2023
Journal article
Open Access
The application of SWAT model and remotely sensed products to characterize the dynamic of streamflow and snow in a mountainous watershed in the High Atlas
Taia S, Erraioui L, Arjdal Y, Chao J, El Mansouri B, Scozzari A
Snowfall, snowpack, and snowmelt are among the processes with the greatest influence on the water cycle in mountainous watersheds. Hydrological models may be significantly biased if snow estimations are inaccurate. However, the unavailability of in situ snow data with enough spatiotemporal resolution limits the application of spatially distributed models in snow-fed watersheds. This obliges numerous modellers to reduce their attention to the snowpack and its effect on water distribution, particularly when a portion of the watershed is predominately covered by snow. This research demonstrates the added value of remotely sensed snow cover products from the Moderate Resolution Imaging Spectroradiometer (MODIS) in evaluating the performance of hydrological models to estimate seasonal snow dynamics and discharge. The Soil and Water Assessment Tool (SWAT) model was used in this work to simulate discharge and snow processes in the Oued El Abid snow-dominated watershed. The model was calibrated and validated on a daily basis, for a long period (1981-2015), using four discharge-gauging stations. A spatially varied approach (snow parameters are varied spatially) and a lumped approach (snow parameters are unique across the whole watershed) have been compared. Remote sensing data provided by MODIS enabled the evaluation of the snow processes simulated by the SWAT model. Results illustrate that SWAT model discharge simulations were satisfactory to good according to the statistical criteria. In addition, the model was able to reasonably estimate the snow-covered area when comparing it to the MODIS daily snow cover product. When allowing snow parameters to vary spatially, SWAT model results were more consistent with the observed streamflow and the MODIS snow-covered area (MODIS-SCA). This paper provides an example of how hydrological modelling using SWAT and snow coverage products by remote sensing may be used together to examine seasonal snow cover and snow dynamics in the High Atlas watershed.Source: SENSORS (BASEL), vol. 23 (issue 3)
DOI: 10.3390/s23031246
Metrics:
Taia S, Erraioui L, Arjdal Y, Chao J, El Mansouri B, Scozzari A
Snowfall, snowpack, and snowmelt are among the processes with the greatest influence on the water cycle in mountainous watersheds. Hydrological models may be significantly biased if snow estimations are inaccurate. However, the unavailability of in situ snow data with enough spatiotemporal resolution limits the application of spatially distributed models in snow-fed watersheds. This obliges numerous modellers to reduce their attention to the snowpack and its effect on water distribution, particularly when a portion of the watershed is predominately covered by snow. This research demonstrates the added value of remotely sensed snow cover products from the Moderate Resolution Imaging Spectroradiometer (MODIS) in evaluating the performance of hydrological models to estimate seasonal snow dynamics and discharge. The Soil and Water Assessment Tool (SWAT) model was used in this work to simulate discharge and snow processes in the Oued El Abid snow-dominated watershed. The model was calibrated and validated on a daily basis, for a long period (1981-2015), using four discharge-gauging stations. A spatially varied approach (snow parameters are varied spatially) and a lumped approach (snow parameters are unique across the whole watershed) have been compared. Remote sensing data provided by MODIS enabled the evaluation of the snow processes simulated by the SWAT model. Results illustrate that SWAT model discharge simulations were satisfactory to good according to the statistical criteria. In addition, the model was able to reasonably estimate the snow-covered area when comparing it to the MODIS daily snow cover product. When allowing snow parameters to vary spatially, SWAT model results were more consistent with the observed streamflow and the MODIS snow-covered area (MODIS-SCA). This paper provides an example of how hydrological modelling using SWAT and snow coverage products by remote sensing may be used together to examine seasonal snow cover and snow dynamics in the High Atlas watershed.Source: SENSORS (BASEL), vol. 23 (issue 3)
DOI: 10.3390/s23031246
Metrics:
See at:
CNR IRIS | ISTI Repository | www.mdpi.com | CNR IRIS
2023
Dataset
Metadata Only Access
A phenotyping weeds image dataset for open scientific research
Dainelli R, Martinelli M, Bruno A, Moroni D, Morelli S, Silvestri M, Ferrari E, Rocchi L, Toscano P
This in-house-built image dataset consists of 10810 weed images captured through a dedicated phenotyping activity in quasi-field conditions. The targets are seven of the most widespread and hard-to-control weeds in wheat (but also in other winter cereals) in the Mediterranean environment. In the framework of open scientific research, our aim is to share low-cost and high-resolution images representing challenging agricultural environments where weather, lighting and other factors can change by the hour and affect the quality of images. This way the dataset could be used to train Artificial Intelligence architectures designed for weed recognition, allowing the implementation of tools directly available in the field for farmers and technicians for effective and timely weed management. The dataset encompasses weed images ranging from the post-emergence phase (i.e. the complete cotyledons unfolding) until the pre-flowering stage. The weed selection was made by considering (i) bottom-up information and specific requests by farmers and technicians, (ii) weed susceptibility to commercial formulations for chemical control <50%, reported at least twice by field technicians, (iii) the difficulty of control considering any methods, and (iv) the type of growing season (overlapping or not with wheat). The final weeds selection encompassed both monocots (Avena sterilis and Lolium multiflorum) and dicots (Convolvulus arvensis, Fumaria officinalis, Papaver rhoeas, Veronica persica and Vicia sativa). Image acquisition was facilitated by using a white panel as a background; this helped to (i) spread the light and thereby make the plants well-illuminated, while still avoiding strong shadows when using the flash and (ii) simplify image processing. The images were acquired with a Canon EOS 700D hand-held camera set in the macro mode with aperture, shutter speed, ISO and flash in auto mode. Photo capture timing, target distances and light conditions did not have a fixed pattern but were deliberately programmed to vary in such a way as to mimic field conditions. For image shooting at various times of the day, the only precaution was to frame the subject with homogeneous light conditions (full sunlight/full shade). The varied outdoor conditions (light, distance, timing) and camera type (RGB) with auto mode were essential features to make the images photos look similar to those that a user can take in a field, for example with a smartphone camera. After selection and categorization, images were cropped to select the region of interest following the 1:1 ratio but maintaining a minimum size of 512 x 512 pixels. More details on the dataset and its use for weed recognition tasks will be soon available in the proceedings of the forthcoming ECPA conference (2-6 July 2023, Bologna, Italy).
Dainelli R, Martinelli M, Bruno A, Moroni D, Morelli S, Silvestri M, Ferrari E, Rocchi L, Toscano P
This in-house-built image dataset consists of 10810 weed images captured through a dedicated phenotyping activity in quasi-field conditions. The targets are seven of the most widespread and hard-to-control weeds in wheat (but also in other winter cereals) in the Mediterranean environment. In the framework of open scientific research, our aim is to share low-cost and high-resolution images representing challenging agricultural environments where weather, lighting and other factors can change by the hour and affect the quality of images. This way the dataset could be used to train Artificial Intelligence architectures designed for weed recognition, allowing the implementation of tools directly available in the field for farmers and technicians for effective and timely weed management. The dataset encompasses weed images ranging from the post-emergence phase (i.e. the complete cotyledons unfolding) until the pre-flowering stage. The weed selection was made by considering (i) bottom-up information and specific requests by farmers and technicians, (ii) weed susceptibility to commercial formulations for chemical control <50%, reported at least twice by field technicians, (iii) the difficulty of control considering any methods, and (iv) the type of growing season (overlapping or not with wheat). The final weeds selection encompassed both monocots (Avena sterilis and Lolium multiflorum) and dicots (Convolvulus arvensis, Fumaria officinalis, Papaver rhoeas, Veronica persica and Vicia sativa). Image acquisition was facilitated by using a white panel as a background; this helped to (i) spread the light and thereby make the plants well-illuminated, while still avoiding strong shadows when using the flash and (ii) simplify image processing. The images were acquired with a Canon EOS 700D hand-held camera set in the macro mode with aperture, shutter speed, ISO and flash in auto mode. Photo capture timing, target distances and light conditions did not have a fixed pattern but were deliberately programmed to vary in such a way as to mimic field conditions. For image shooting at various times of the day, the only precaution was to frame the subject with homogeneous light conditions (full sunlight/full shade). The varied outdoor conditions (light, distance, timing) and camera type (RGB) with auto mode were essential features to make the images photos look similar to those that a user can take in a field, for example with a smartphone camera. After selection and categorization, images were cropped to select the region of interest following the 1:1 ratio but maintaining a minimum size of 512 x 512 pixels. More details on the dataset and its use for weed recognition tasks will be soon available in the proceedings of the forthcoming ECPA conference (2-6 July 2023, Bologna, Italy).
See at:
CNR IRIS | zenodo.org
2023
Conference article
Restricted
Recognition of weeds in cereals using AI architecture
Dainelli R, Martinelli M, Bruno A, Moroni D, Morelli S, Silvestri M, Ferrari E, Rocchi L, Toscano P
In this study, an automatic system based on open AI architectures was developed and fed with an in-house built image dataset to recognize seven of the most widespread and hard-to-control weeds in wheat in the Mediterranean environment. A total of 10810 images were collected from the post-emergence (S1 dataset) to the pre-flowering stage (S2 dataset). A selection of pictures available from online sources (S3, 825 images) was used as a final and further independent test of the proposed recognition tool. The AI tool in the ensemble configuration achieved 100% accuracy on the validation and test set both for S1 and S2, while for S3 an accuracy of approximately 70% was achieved for weed species in the post-emergence stage.DOI: 10.3920/978-90-8686-947-3_49
Metrics:
Dainelli R, Martinelli M, Bruno A, Moroni D, Morelli S, Silvestri M, Ferrari E, Rocchi L, Toscano P
In this study, an automatic system based on open AI architectures was developed and fed with an in-house built image dataset to recognize seven of the most widespread and hard-to-control weeds in wheat in the Mediterranean environment. A total of 10810 images were collected from the post-emergence (S1 dataset) to the pre-flowering stage (S2 dataset). A selection of pictures available from online sources (S3, 825 images) was used as a final and further independent test of the proposed recognition tool. The AI tool in the ensemble configuration achieved 100% accuracy on the validation and test set both for S1 and S2, while for S3 an accuracy of approximately 70% was achieved for weed species in the post-emergence stage.DOI: 10.3920/978-90-8686-947-3_49
Metrics:
See at:
doi.org | CNR IRIS | CNR IRIS | www.wageningenacademic.com
2023
Journal article
Open Access
MEC: a Mesoscale Events Classifier for oceanographic imagery
Pieri G, Janeiro J, Martins F, Papini O, Reggiannini M
The observation of the sea through remote sensing technologies plays a fundamental role in understanding the state of health of marine fauna species and their behaviour. Mesoscale phenomena, such as upwelling, countercurrents, and filaments, are essential processes to be analysed because their occurrence involves, among other things, variations in the density of nutrients, which, in turn, influence the biological parameters of the habitat. Indeed, there is a connection between the biogeochemical and physical processes that occur within a biological system and the variations observed in its faunal populations. This paper concerns the proposal of an automatic classification system, namely the Mesoscale Events Classifier, dedicated to the recognition of marine mesoscale events. The proposed system is devoted to the study of these phenomena through the analysis of sea surface temperature images captured by satellite missions, such as EUMETSAT's Metop and NASA's Earth Observing System programmes. The classification of these images is obtained through (i) a preprocessing stage with the goal to provide a simultaneous representation of the spatial and temporal properties of the data and enhance the salient features of the sought phenomena, (ii) the extraction of temporal and spatial characteristics from the data and, finally, (iii) the application of a set of rules to discriminate between different observed scenarios. The results presented in this work were obtained by applying the proposed approach to images acquired in the southwestern region of the Iberian peninsula.Source: APPLIED SCIENCES, vol. 13 (issue 3)
DOI: 10.3390/app13031565
Project(s): NAUTILOS
Metrics:
Pieri G, Janeiro J, Martins F, Papini O, Reggiannini M
The observation of the sea through remote sensing technologies plays a fundamental role in understanding the state of health of marine fauna species and their behaviour. Mesoscale phenomena, such as upwelling, countercurrents, and filaments, are essential processes to be analysed because their occurrence involves, among other things, variations in the density of nutrients, which, in turn, influence the biological parameters of the habitat. Indeed, there is a connection between the biogeochemical and physical processes that occur within a biological system and the variations observed in its faunal populations. This paper concerns the proposal of an automatic classification system, namely the Mesoscale Events Classifier, dedicated to the recognition of marine mesoscale events. The proposed system is devoted to the study of these phenomena through the analysis of sea surface temperature images captured by satellite missions, such as EUMETSAT's Metop and NASA's Earth Observing System programmes. The classification of these images is obtained through (i) a preprocessing stage with the goal to provide a simultaneous representation of the spatial and temporal properties of the data and enhance the salient features of the sought phenomena, (ii) the extraction of temporal and spatial characteristics from the data and, finally, (iii) the application of a set of rules to discriminate between different observed scenarios. The results presented in this work were obtained by applying the proposed approach to images acquired in the southwestern region of the Iberian peninsula.Source: APPLIED SCIENCES, vol. 13 (issue 3)
DOI: 10.3390/app13031565
Project(s): NAUTILOS
Metrics:
See at:
CNR IRIS | ISTI Repository | www.mdpi.com | CNR IRIS
2023
Other
Restricted
Relazione conclusiva attività svolte nel periodo Gennaio 2020 - Gennario 2023. Accordo CNR-ISTI SIMeM
Martinelli M, Pratali L
Relazione conclusiva attività svolte nel periodo Gennaio 2020 - Gennario 2023. Accordo CNR-ISTI SIMeM
Martinelli M, Pratali L
Relazione conclusiva attività svolte nel periodo Gennaio 2020 - Gennario 2023. Accordo CNR-ISTI SIMeM
2023
Journal article
Open Access
Exploiting the haptic and audio channels to improve orientation and mobility apps for the visually impaired
Paratore Mt, Leporini B
Orientation and mobility apps for visually impaired people are well known to be effective in improving the quality of life for this target group. A mobile application that guides a visually impaired person step-by-step through a physical space is a valuable aid, but it does not provide an overview of a complex environment "at a glance," as a traditional hard-copy tactile map does. The aim of this study is to investigate whether a smartphone GPS map, enriched with haptic and audio hints, can facilitate cognitive mapping for visually impaired users. Encouraged by a preliminary study conducted in co-operation with two visually impaired volunteers, we designed and developed an Android prototype for exploration of an urban area. Our goal was to provide an affordable, portable and versatile solution to help users increase awareness of an environment through the positions of its landmarks and points of interest. Vibro-tactile and audio hints were linked to the coordinates on the map via the GeoJSON data format and were issued exploiting the text-to-speech and vibration features of the mobile device, as they were displayed through the operating system's APIs. Test sessions and interviews with visually impaired users produced encouraging results. Results, to be verified by more extensive testing, overall confirm the validity of our approach and are in line with results found in the literature.Source: UNIVERSAL ACCESS IN THE INFORMATION SOCIETY (PRINT)
DOI: 10.1007/s10209-023-00973-4
Metrics:
Paratore Mt, Leporini B
Orientation and mobility apps for visually impaired people are well known to be effective in improving the quality of life for this target group. A mobile application that guides a visually impaired person step-by-step through a physical space is a valuable aid, but it does not provide an overview of a complex environment "at a glance," as a traditional hard-copy tactile map does. The aim of this study is to investigate whether a smartphone GPS map, enriched with haptic and audio hints, can facilitate cognitive mapping for visually impaired users. Encouraged by a preliminary study conducted in co-operation with two visually impaired volunteers, we designed and developed an Android prototype for exploration of an urban area. Our goal was to provide an affordable, portable and versatile solution to help users increase awareness of an environment through the positions of its landmarks and points of interest. Vibro-tactile and audio hints were linked to the coordinates on the map via the GeoJSON data format and were issued exploiting the text-to-speech and vibration features of the mobile device, as they were displayed through the operating system's APIs. Test sessions and interviews with visually impaired users produced encouraging results. Results, to be verified by more extensive testing, overall confirm the validity of our approach and are in line with results found in the literature.Source: UNIVERSAL ACCESS IN THE INFORMATION SOCIETY (PRINT)
DOI: 10.1007/s10209-023-00973-4
Metrics:
See at:
CNR IRIS | link.springer.com | ISTI Repository | CNR IRIS | CNR IRIS
2023
Other
Restricted
Artificial intelligence for image classification of wheat's phytopathologies, pests and infestants
Bruno A, Martinelli M, Moroni D, Rocchi L, Morelli S, Ferrari E, Toscano P, Dainelli R
In this work we present a solution to deal with issues related to wheat cultures: diseases, abiotic damages, weeds and pests. The core of this work is the EfficientNet architecture, in particular the b0 version which is the one with best accuracy/complexity ratio in our opinion. Results show that in this way accuracies are improved by +2% up to +10% while the complexity is unchanged. The models presented and tested during this work have been deployed and can be tested by using the mobile app Granoscan.Source: COMPUTERS AND ELECTRONICS IN AGRICULTURE
Bruno A, Martinelli M, Moroni D, Rocchi L, Morelli S, Ferrari E, Toscano P, Dainelli R
In this work we present a solution to deal with issues related to wheat cultures: diseases, abiotic damages, weeds and pests. The core of this work is the EfficientNet architecture, in particular the b0 version which is the one with best accuracy/complexity ratio in our opinion. Results show that in this way accuracies are improved by +2% up to +10% while the complexity is unchanged. The models presented and tested during this work have been deployed and can be tested by using the mobile app Granoscan.Source: COMPUTERS AND ELECTRONICS IN AGRICULTURE
2023
Conference article
Open Access
Medical waste sorting: a computer vision approach for assisted primary sorting
Bruno A, Caudai C, Leone Gr, Martinelli M, Moroni D, Crotti F
Medical waste, i.e. waste produced during medical activities in hospitals, clinics and laboratories, represents hazardous waste whose management requires special care and high costs. However, this kind of waste contains a large fraction of highly valued materials that can enter a circular economy process. To this end, in this paper, we propose a computer vision approach for assisting in the primary sorting of med- ical waste. The feasibility of our approach is demonstrated on representative datasets we collected and made available to the community.DOI: 10.1109/icasspw59220.2023.10193520
Metrics:
Bruno A, Caudai C, Leone Gr, Martinelli M, Moroni D, Crotti F
Medical waste, i.e. waste produced during medical activities in hospitals, clinics and laboratories, represents hazardous waste whose management requires special care and high costs. However, this kind of waste contains a large fraction of highly valued materials that can enter a circular economy process. To this end, in this paper, we propose a computer vision approach for assisting in the primary sorting of med- ical waste. The feasibility of our approach is demonstrated on representative datasets we collected and made available to the community.DOI: 10.1109/icasspw59220.2023.10193520
Metrics:
See at:
CNR IRIS | ieeexplore.ieee.org | ISTI Repository | CNR IRIS | CNR IRIS
2023
Other
Open Access
Working Group on Nephrops Surveys (WGNEPS; outputs from 2022 meeting)
Aguzzi J, Aristeguiezquibela M, Burgos C, Chatzievangelou D, Doyle J, Fallon N, Fifas S, Gonzálezherraiz I, Jonsson P, Lundy M, Martinelli M, Medveek D, Naseer A, Nava E, Nawri N, Jónasson Jp, Pereira B, Pieri G, Silva C, Tibone M, Valeiras J, Vila Y, Weetman A, Wieland K
The Working Group on Nephrops Surveys (WGNEPS) is the international coordination group for Nephrops underwater television and trawl surveys within ICES. This report summarizes the na-tional contributions on the results of the surveys conducted in 2022 together with time series covering all survey years, problems encountered, data quality checks and technological improve-ments as well as the planning for survey activities for 2023. In total, 21 surveys covering 26 functional units (FU's) in the ICES area and 1 geographical sub-area (GSA) in the Adriatic Sea were discussed and further improvements in respect to survey design and data analysis standardization and the use of most recent technology were reviewed. The first exploratory UWTV survey on the FU 25 Nephrops grounds was also presented to the group. The results of the evaluation of reference sets for FU3&4 Skagerrak/Kattegat were accepted fol-lowing the process set down by the 2018 workshop (WKNEPS). An alternative method estimate Nephrops abundance was shown to the group using the recently published R package sdmTMB. The group agreed to hold a workshop in 2025 to address burrow size estimations to update cor-rection factors and terms of reference for this to be agreed at next meeting. Automatic burrow detection based on deep learning methods continues to show promising re-sults where datasets from multiple institutes were used. Plans are being progressed for an international Nephrops UWTV database to be established at the ICES data centre with a sub-group.Source: ICES SCIENTIFIC REPORTS
DOI: 10.17895/ices.pub.22211161
Metrics:
Aguzzi J, Aristeguiezquibela M, Burgos C, Chatzievangelou D, Doyle J, Fallon N, Fifas S, Gonzálezherraiz I, Jonsson P, Lundy M, Martinelli M, Medveek D, Naseer A, Nava E, Nawri N, Jónasson Jp, Pereira B, Pieri G, Silva C, Tibone M, Valeiras J, Vila Y, Weetman A, Wieland K
The Working Group on Nephrops Surveys (WGNEPS) is the international coordination group for Nephrops underwater television and trawl surveys within ICES. This report summarizes the na-tional contributions on the results of the surveys conducted in 2022 together with time series covering all survey years, problems encountered, data quality checks and technological improve-ments as well as the planning for survey activities for 2023. In total, 21 surveys covering 26 functional units (FU's) in the ICES area and 1 geographical sub-area (GSA) in the Adriatic Sea were discussed and further improvements in respect to survey design and data analysis standardization and the use of most recent technology were reviewed. The first exploratory UWTV survey on the FU 25 Nephrops grounds was also presented to the group. The results of the evaluation of reference sets for FU3&4 Skagerrak/Kattegat were accepted fol-lowing the process set down by the 2018 workshop (WKNEPS). An alternative method estimate Nephrops abundance was shown to the group using the recently published R package sdmTMB. The group agreed to hold a workshop in 2025 to address burrow size estimations to update cor-rection factors and terms of reference for this to be agreed at next meeting. Automatic burrow detection based on deep learning methods continues to show promising re-sults where datasets from multiple institutes were used. Plans are being progressed for an international Nephrops UWTV database to be established at the ICES data centre with a sub-group.Source: ICES SCIENTIFIC REPORTS
DOI: 10.17895/ices.pub.22211161
Metrics:
See at:
CNR IRIS | ISTI Repository | CNR IRIS
2023
Contribution to book
Open Access
Italy and Croatia : Pomo Pits, Central Adriatic Sea (GSA 17) Adriatic UWTV Surveys and Pomo monitoring activity
Martinelli M, Medveek D, Domenichetti F, Canduci G, Giuliani G, Zacchetti L, Pieri G, Belardinelli A, Chiarini M, Guicciardi S, Grilli F, Penna P, Scarpini P, Cvitani R, Isajlovic I, Vrgoc N
The Pomo (or Jabuka) Pits area is one of the main fishing ground for Norway Lobster Nephrops norvegicus and European hake Merluccius merluccius within the GFCM Geographical Sub Areas 17 (Northern and Central Adriatic Sea) and it is shared by the Italian and the Croatian fleets. Furthermore, this represents a well-known nursery area for M. merluccius and hosts a distinct population of N. norvegicus, characterized by small-sized mature individuals. Due to a decline in landing of both species for the Adriatic Sea, since 2015 the Italian and the Croatian governments implemented some protection measures in that area. Eventually in 2018, the GFCM established a Fishery Restricted Area. Since 2009 the area is yearly monitored by CNR IRBIM in collaboration with IOF Split. From 2009 to 2019 (except 2011 and 2018), a spring UWTV survey was conducted in the Pomo Pits area jointly by CNR-IRBIM Ancona and IOF Split, on board the CNR R/V Dallaporta; the Pomo Pits UWTV time series has been recently included, as a tuning index, in new modeling approaches tested for the Adriatic N. norvegicus stock assessment. Trials on automatic burrow tracking and counting have also been recently conducted on the Adriatic UWTV footage in the framework of the EU H2020 NAUTILOS project.Source: ICES SCIENTIFIC REPORTS, pp. 119-123
DOI: 10.17895/ices.pub.22211161
Project(s): NAUTILOS
Metrics:
Martinelli M, Medveek D, Domenichetti F, Canduci G, Giuliani G, Zacchetti L, Pieri G, Belardinelli A, Chiarini M, Guicciardi S, Grilli F, Penna P, Scarpini P, Cvitani R, Isajlovic I, Vrgoc N
The Pomo (or Jabuka) Pits area is one of the main fishing ground for Norway Lobster Nephrops norvegicus and European hake Merluccius merluccius within the GFCM Geographical Sub Areas 17 (Northern and Central Adriatic Sea) and it is shared by the Italian and the Croatian fleets. Furthermore, this represents a well-known nursery area for M. merluccius and hosts a distinct population of N. norvegicus, characterized by small-sized mature individuals. Due to a decline in landing of both species for the Adriatic Sea, since 2015 the Italian and the Croatian governments implemented some protection measures in that area. Eventually in 2018, the GFCM established a Fishery Restricted Area. Since 2009 the area is yearly monitored by CNR IRBIM in collaboration with IOF Split. From 2009 to 2019 (except 2011 and 2018), a spring UWTV survey was conducted in the Pomo Pits area jointly by CNR-IRBIM Ancona and IOF Split, on board the CNR R/V Dallaporta; the Pomo Pits UWTV time series has been recently included, as a tuning index, in new modeling approaches tested for the Adriatic N. norvegicus stock assessment. Trials on automatic burrow tracking and counting have also been recently conducted on the Adriatic UWTV footage in the framework of the EU H2020 NAUTILOS project.Source: ICES SCIENTIFIC REPORTS, pp. 119-123
DOI: 10.17895/ices.pub.22211161
Project(s): NAUTILOS
Metrics:
See at:
CNR IRIS | ISTI Repository | CNR IRIS
2023
Journal article
Open Access
Efficient lung ultrasound classification
Bruno A, Ignesti G, Salvetti O, Moroni D, Martinelli M
A machine learning method for classifying Lung UltraSound is here proposed to provide a point of care tool for supporting a safe, fast and accurate diagnosis, that can also be useful during a pandemic like as SARS-CoV-2. Given the advantages (e.g. safety, rapidity, portability, cost-effectiveness) provided by the ultrasound technology over other methods (e.g. X-ray, computer tomography, magnetic resonance imaging), our method was validated on the largest LUS public dataset. Focusing on both accuracy and efficiency, our solution is based on an efficient adaptive ensembling of two EfficientNet-b0 models reaching 100% of accuracy, which, to our knowledge, outperforms the previous state-of-the-art. The complexity of this solution keeps the number of parameters in the same order as an EfficientNet-b0 by adopting specific design choices that are adaptive ensembling with a combination layer, ensembling performed on the deep features, minimal ensemble only two weak models. Moreover, a visual analysis of the saliency maps on sample images of all the classes of the dataset reveals where the focus is on an inaccurate weak model versus an accurate model.Source: BIOENGINEERING, vol. 10 (issue 5)
DOI: 10.3390/bioengineering10050555
Metrics:
Bruno A, Ignesti G, Salvetti O, Moroni D, Martinelli M
A machine learning method for classifying Lung UltraSound is here proposed to provide a point of care tool for supporting a safe, fast and accurate diagnosis, that can also be useful during a pandemic like as SARS-CoV-2. Given the advantages (e.g. safety, rapidity, portability, cost-effectiveness) provided by the ultrasound technology over other methods (e.g. X-ray, computer tomography, magnetic resonance imaging), our method was validated on the largest LUS public dataset. Focusing on both accuracy and efficiency, our solution is based on an efficient adaptive ensembling of two EfficientNet-b0 models reaching 100% of accuracy, which, to our knowledge, outperforms the previous state-of-the-art. The complexity of this solution keeps the number of parameters in the same order as an EfficientNet-b0 by adopting specific design choices that are adaptive ensembling with a combination layer, ensembling performed on the deep features, minimal ensemble only two weak models. Moreover, a visual analysis of the saliency maps on sample images of all the classes of the dataset reveals where the focus is on an inaccurate weak model versus an accurate model.Source: BIOENGINEERING, vol. 10 (issue 5)
DOI: 10.3390/bioengineering10050555
Metrics:
See at:
CNR IRIS | ISTI Repository | www.mdpi.com | CNR IRIS | CNR IRIS