Method for Segmentation of Litchi Branches Based on the Improved DeepLabv3+

Abstract

It is necessary to develop automatic picking technology to improve the efficiency of litchi picking, and the accurate segmentation of litchi branches is the key that allows robots to complete the picking task. To solve the problem of inaccurate segmentation of litchi branches under natural conditions, this paper proposes a segmentation method for litchi branches based on the improved DeepLabv3+, which replaced the backbone network of DeepLabv3+ and used the Dilated Residual Networks as the backbone network to enhance the model’s feature extraction capability. During the training process, a combination of Cross-Entropy loss and the dice coefficient loss was used as the loss function to cause the model to pay more attention to the litchi branch area, which could alleviate the negative impact of the imbalance between the litchi branches and the background. In addition, the Coordinate Attention module is added to the atrous spatial pyramid pooling, and the channel and location information of the multi-scale semantic features acquired by the network are simultaneously considered. The experimental results show that the model’s mean intersection over union and mean pixel accuracy are 90.28% and 94.95%, respectively, and the frames per second (FPS) is 19.83. Compared with the classical DeepLabv3+ network, the model’s mean intersection over union and mean pixel accuracy are improved by 13.57% and 15.78%, respectively. This method can accurately segment litchi branches, which provides powerful technical support to help litchi-picking robots find branches.

STAE‐YOLO: Intelligent detection algorithm for risk management of construction machinery intrusion on transmission lines based on visual perception

“…To comprehensively assess the performance of the models, we selected several state‐of‐the‐art (SOTA) object detection networks for further comparison, each representing the most advanced algorithms in the current field. These include the two‐stage detection network Faster R‐CNN [39], the single‐stage detection network RetinaNet [45], SSD [42], YOLOv5 [40], YOLOv8 [41], the single‐stage anchor‐free network CenterNet [44], the end‐to‐end detection network DETR [43], and the single‐stage anchor‐free network FCOS [46]. Additionally, we replaced our improvements on the commonly used YOLOv5‐S model, incorporating the YOLOv5s‐Swin Transformer and YOLOv5s‐SA as two comparative enhanced models, and validated all obtained results.…”

A Real-Time Dual-Task Defect Segmentation Network for Grinding Wheels with Coordinate Attentioned-ASP and Masked Autoencoder

“…One enlightenment brought by the Self-Attentions is that global information can be incorporated into the features, not only through concatenation [21][22][23][24][25][26] or addition [14,20], but also through weighted multiplication [6,7,[46][47][48]. Separable Self-Attention is a classic structure of Self-Attentions, and CA is a typical case of attention mechanisms.…”

“…Liu et al [17] presented a novel Dual Value Attention (DVA) module with a pair of ASPP modules, followed by a self-attention module in the Panoptic DeepLab. Moreover, the combination of ASPP and the Convolutional Block Attention Module (CBAM) [18] or Coordinate Attention (CA) [19] is frequently adopted to solve respective tasks [20][21][22][23][24][25][26]. Similar approaches can be found in other architectures like the U-Net [27][28][29][30][31].…”

“…Embedded into the multi-scale contextual features extractor like ASPP, to further highlight multi-scale information interactions [15,17,[20][21][22][23][24][25][26];…”

Prototype Network for Predicting Occluded Picking Position Based on Lychee Phenotypic Features

“…Some studies used DeepLabV3 to segment and identify litchi fruits or branches [3][4][5], and some studies used yolo technology to detect and identify litchi [6][7][8][9]. For lychee clusters with stem diameters obscured from leaf occlusion, it is challenging to locate them based on texture in RGB images [10][11][12][13]. Neural networks struggle to learn positional features for occluded lychee clusters.…”