The ripeness of bananas is the most significant factor affecting nutrient composition and demand. Conventionally, cutting and ripeness analysis requires expert knowledge and substantial human intervention, and different studies have been conducted to automate and substantially reduce human effort. Using the Preferred Reporting Items for the Systematic Reviews approach, 1548 studies were extracted from journals and conferences, using different research databases, and 35 were included in the final review for key parameters. These studies suggest the dominance of banana fingers as input data, a sensor camera as the preferred capturing device, and appropriate features, such as color, that can provide better detection. Among six stages of ripeness, the studies employing the four mentioned stages performed better in terms of accuracy and coefficient of determination value. Among all the works for detecting ripeness stages prediction, convolutional neural networks were found to perform sufficiently well with large datasets, whereas conventional artificial neural networks and support vector machines attained better performance for sensor-related data. However, insufficient information on the dataset and capturing device, limited data availability, and exploitation of data augmentation techniques are limitations in existing studies. Thus, effectively addressing these shortcomings and close collaboration with experts to predict the ripeness stages should be pursued.

Verbal and non-verbal activities convey insightful information about people’s affect, empathy, and engagement during social interactions. In this paper, we investigate the usage of inertial sensors to recognize verbal (e.g., speaking), non-verbal (e.g., head nodding, shaking) and other activities (e.g., eating, no movement). We implement an end-to-end deep neural network to distinguish among these activities. We then explore the generalizability of the approach in three scenarios: (1) using new data to detect a known activity from a known user, (2) detecting a novel activity of a known user and (3) detecting the activity of an unknown user. Results show that using accelerometer and gyroscope sensors, the model achieves a balanced accuracy of 55% when tested on data from a new user, 41% on a new activity of an existing user, and 80% on new data of a known activity from an existing user. The results are between 7-47 percentage points higher than baseline classifiers.

The number of coronavirus disease (COVID-19) cases is constantly rising as the pandemic continues, with new variants constantly emerging. Therefore, to prevent the virus from spreading, coronavirus cases must be diagnosed as soon as possible. The COVID-19 pandemic has had a devastating impact on people’s health and the economy worldwide. For COVID-19 detection, reverse transcription-polymerase chain reaction testing is the benchmark. However, this test takes a long time and necessitates a lot of laboratory resources. A new trend is emerging to address these limitations regarding the use of machine learning and deep learning techniques for automatic analysis, as these can attain high diagnosis results, especially by using medical imaging techniques. However, a key question arises whether a chest computed tomography scan or chest X-ray can be used for COVID-19 detection. A total of 17,599 images were examined in this work to develop the models used to classify the occurrence of COVID-19 infection, while four different classifiers were studied. These are the convolutional neural network (proposed architecture (named, SCovNet) and Resnet18), support vector machine, and logistic regression. Out of all four models, the proposed SCoVNet architecture reached the best performance with an accuracy of almost 99% and 98% on chest computed tomography scan images and chest X-ray images, respectively.

The conventional process of visual detection and manual harvesting of the banana bunch has been a known problem faced by the agricultural industry. It is a laborious activity associated with inconsistency in the inspection and grading process, leading to post-harvest losses. Automated fruit harvesting using computer vision empowered by deep learning could significantly impact the visual inspection process domains, allowing consistent harvesting and grading. To achieve the goal of the industry-level harvesting process, this work collects data from professional harvesters from the industry. It investigates six state-of-the-art architectures to find the best solution. 2,685 samples were collected from four different sites with expert opinions from industry harvesters to cut (or harvest) and keep (or not harvest) the banana brunch. Comparative results showed that the DenseNet121 architecture outperformed the other examined architectures, reaching a precision, recall, F1 score, accuracy, and specificity of 85%, 82%, 82%, 83%, and 83%, respectively. In addition, an understanding of the underlying black box nature of the solution was visualized and found adequate. This visual interpretation of the model supports human expert’s criteria for harvesting. This system can assist or replace human experts in the field.

This study introduces a multi-part dataset designed to support the development of artificial intelligence models for banana bunch detection and harvesting decision-making. The dataset includes images collected from four different fields in Madeira Island, Portugal, under varying environmental conditions. It is divided into three core subsets, namely, a detection dataset annotated using You Only Look Once (YOLO) format (2179 images labelled for bunch and flower bud detection), a harvesting classification dataset labelled by expert teams as “Cut”“ and “Keep” (2685 images, with 1143 labelled as “Cut” and 1542 as “Keep”) and an expert opinion dataset where images were classified by human experts into three decision categories: “Cut now”, “Keep for next cut” and “Wait more than three weeks” (400 images, with 100 samples evaluated by each of four expert cutters, capturing all three decision categories) These datasets enable the creation and benchmarking of computer vision models and allow for expert consensus analysis.