FruitSeg30_Segmentation dataset & mask annotations: A novel dataset for diverse fruit segmentation and classification.

Fruits are mature ovaries of flowering plants that are integral to human diets, providing essential nutrients such as vitamins, minerals, fiber and antioxidants that are crucial for health and disease prevention. Accurate classification and segmentation of fruits are crucial in the agricultural sector for enhancing the efficiency of sorting and quality control processes, which significantly benefit automated systems by reducing labor costs and improving product consistency. This paper introduces the "FruitSeg30_Segmentation Dataset & Mask Annotations", a novel dataset designed to advance the capability of deep learning models in fruit segmentation and classification. Comprising 1969 high-quality images across 30 distinct fruit classes, this dataset provides diverse visuals essential for a robust model. Utilizing a U-Net architecture, the model trained on this dataset achieved training accuracy of 94.72 %, validation accuracy of 92.57 %, precision of 94 %, recall of 91 %, f1-score of 92.5 %, IoU score of 86 %, and maximum dice score of 0.9472, demonstrating superior performance in segmentation tasks. The FruitSeg30 dataset fills a critical gap and sets new standards in dataset quality and diversity, enhancing agricultural technology and food industry applications.

Ripening dynamics revisited: an automated method to track the development of asynchronous berries on time-lapse images.

BACKGROUND: Grapevine berries undergo asynchronous growth and ripening dynamics within the same bunch. Due to the lack of efficient methods to perform sequential non-destructive measurements on a representative number of individual berries, the genetic and environmental origins of this heterogeneity, remain nearly unknown. To address these limitations, we propose a method to track the growth and coloration kinetics of individual berries on time-lapse images of grapevine bunches. RESULTS: First, a deep-learning approach is used to detect berries with at least 50 ± 10% of visible contours, and infer the shape they would have in the absence of occlusions. Second, a tracking algorithm was developed to assign a common label to shapes representing the same berry along the time-series. Training and validation of the methods were performed on challenging image datasets acquired in a robotised high-throughput phenotyping platform. Berries were detected on various genotypes with a F1-score of 91.8%, and segmented with a mean absolute error of 4.1% on their area. Tracking allowed to label and retrieve the temporal identity of more than half of the segmented berries, with an accuracy of 98.1%. This method was used to extract individual growth and colour kinetics of various berries from the same bunch, allowing us to propose the first statistically relevant analysis of berry ripening kinetics, with a time resolution lower than one day. CONCLUSIONS: We successfully developed a fully-automated open-source method to detect, segment and track overlapping berries in time-series of grapevine bunch images acquired in laboratory conditions. This makes it possible to quantify fine aspects of individual berry development, and to characterise the asynchrony within the bunch. The interest of such analysis was illustrated here for one cultivar, but the method has the potential to be applied in a high throughput phenotyping context. This opens the way for revisiting the genetic and environmental variations of the ripening dynamics. Such variations could be considered both from the point of view of fruit development and the phenological structure of the population, which would constitute a paradigm shift.

DAM: Hierarchical Adaptive Feature Selection Using Convolution Encoder Decoder Network for Strawberry Segmentation.

Autonomous harvesters can be used for the timely cultivation of high-value crops such as strawberries, where the robots have the capability to identify ripe and unripe crops. However, the real-time segmentation of strawberries in an unbridled farming environment is a challenging task due to fruit occlusion by multiple trusses, stems, and leaves. In this work, we propose a possible solution by constructing a dynamic feature selection mechanism for convolutional neural networks (CNN). The proposed building block namely a dense attention module (DAM) controls the flow of information between the convolutional encoder and decoder. DAM enables hierarchical adaptive feature fusion by exploiting both inter-channel and intra-channel relationships and can be easily integrated into any existing CNN to obtain category-specific feature maps. We validate our attention module through extensive ablation experiments. In addition, a dataset is collected from different strawberry farms and divided into four classes corresponding to different maturity levels of fruits and one is devoted to background. Quantitative analysis of the proposed method showed a 4.1% and 2.32% increase in mean intersection over union, over existing state-of-the-art semantic segmentation models and other attention modules respectively, while simultaneously retaining a processing speed of 53 frames per second.