RVDR-YOLOv8: A Weed Target Detection Model Based on Improved YOLOv8

Yuanming Ding,Chen Jiang,Lin Song,Fei Liu,Yunrui Tao

DOI: https://doi.org/10.3390/electronics13112182

IF: 2.9

2024-06-04

Electronics

Abstract:Currently, weed control robots that can accurately identify weeds and carry out removal work are gradually replacing traditional chemical weed control techniques. However, the computational and storage resources of the core processing equipment of weeding robots are limited. Aiming at the current problems of high computation and the high number of model parameters in weeding robots, this paper proposes a lightweight weed target detection model based on the improved YOLOv8 (You Only Look Once Version 8), called RVDR-YOLOv8 (Reversible Column Dilation-wise Residual). First, the backbone network is reconstructed based on RevCol (Reversible Column Networks). The unique reversible columnar structure of the new backbone network not only reduces the computational volume but also improves the model generalisation ability. Second, the C2fDWR module is designed using Dilation-wise Residual and integrated with the reconstructed backbone network, which improves the adaptive ability of the new backbone network RVDR and enhances the model's recognition accuracy for occluded targets. Again, GSConv is introduced at the neck end instead of traditional convolution to reduce the complexity of computation and network structure while ensuring the model recognition accuracy. Finally, InnerMPDIoU is designed by combining MPDIoU with InnerIoU to improve the prediction accuracy of the model. The experimental results show that the computational complexity of the new model is reduced by 35.8%, the number of parameters is reduced by 35.4% and the model size is reduced by 30.2%, while the mAP50 and mAP50-95 values are improved by 1.7% and 1.1%, respectively, compared to YOLOv8. The overall performance of the new model is improved compared to models such as Faster R-CNN, SSD and RetinaNet. The new model proposed in this paper can achieve the accurate identification of weeds in farmland under the condition of limited hardware resources, which provides theoretical and technical support for the effective control of weeds in farmland.

engineering, electrical & electronic,computer science, information systems,physics, applied

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to achieve efficient and accurate weed target detection in weed - control robots. Specifically, although current weed - control robots can gradually replace traditional chemical weeding techniques, the computing and storage resources of their core processing devices are limited. Therefore, the paper proposes a lightweight weed target detection model based on the improved YOLOv8, named RVDR - YOLOv8 (Reversible Column Dilation - wise Residual). This model aims to reduce the computational complexity and the number of model parameters while improving the recognition accuracy. In particular, it can achieve accurate identification of weeds in farmland under hardware - constrained conditions.

### Main problems solved in the paper:
1. **High computational complexity**: Traditional weed detection models are computationally intensive, which limits their application on robots with limited resources.
2. **A large number of model parameters**: The existing models have a large number of parameters, resulting in a large model volume and occupying a large amount of storage space, which is not conducive to the deployment of embedded devices.
3. **Insufficient ability to detect small targets**: The existing YOLO series models have certain defects in detecting small targets, especially in complex farmland environments.

### Solutions:
1. **Reconstruct the backbone network**: Use RevCol (Reversible Column Networks) to reconstruct the backbone network of YOLOv8 to reduce the amount of computation and improve the generalization ability of the model.
2. **Design the C2fDWR module**: Combine Dilation - wise Residual to design the C2fDWR module to enhance the model's ability to recognize occluded targets.
3. **Introduce GSConv**: Use GSConv instead of traditional convolution in the neck of the model to reduce the computational complexity and the complexity of the network structure while maintaining the recognition accuracy of the model.
4. **Optimize the loss function**: Design the InnerMPDIoU loss function, combining MPDIoU and InnerIoU, to improve the prediction accuracy of the model.

Through the above improvements, the RVDR - YOLOv8 model proposed in the paper has been reduced in terms of computational complexity, number of parameters, and model size, while also being improved in the mAP 50 and mAP 50 - 95 indicators. Its overall performance is better than other models, such as Faster R - CNN, SSD, and RetinaNet. These improvements enable the model to achieve efficient weed identification on weed - control robots with limited resources.