# WormNet: A Multi-View Network for Silkworm Re-Identification

## TL;DR

This paper introduces WormNet, a new AI method for identifying individual silkworms using multi-view networks, which could help improve silkworm farming practices.

## Contribution

The novel WormNet architecture introduces multi-order feature extraction, spatial purification, and channel interaction for silkworm re-identification.

## Key findings

- WormNet achieves an mAP of 54.8% and rank-1 accuracy of 91.4% on a new silkworm ReID dataset.
- The proposed network outperforms existing person and object ReID models in silkworm identification.
- The study introduces a new dataset for silkworm ReID training and evaluation.

## Abstract

In this study, we explore the application of artificial intelligence in silkworm farming by integrating re-identification (ReID) techniques with individual silkworm recognition and propose a novel silkworm ReID method. To address the challenges posed by arbitrary silkworm postures, high inter-individual similarity, and complex image backgrounds—which hinder the performance of existing pedestrian or object ReID models on silkworm datasets—a specialized network architecture is developed. This architecture incorporates multi-order feature extraction, spatial purification, and channel interaction mechanisms to enhance the network’s performance in identifying individual silkworms. The experimental results demonstrate that the proposed network outperforms baseline networks, attention-based models, and state-of-the-art person ReID approaches. This research provides a valuable reference for individual identification in animals and insects.

Re-identification (ReID) has been widely applied in person and vehicle recognition tasks. This study extends its application to a novel domain: insect (silkworm) recognition. However, unlike person or vehicle ReID, silkworm ReID presents unique challenges, such as the high similarity between individuals, arbitrary poses, and significant background noise. To address these challenges, we propose a multi-view network for silkworm ReID, called WormNet, which is built upon an innovative strategy termed extraction purification extraction interaction. Specifically, we introduce a multi-order feature extraction module that captures a wide range of fine-grained features by utilizing convolutional kernels of varying sizes and parallel cardinality, effectively mitigating issues of high individual similarity and diverse poses. Next, a feature mask module (FMM) is employed to purify the features in the spatial domain, thereby reducing the impact of background interference. To further enhance the data representation capabilities of the network, we propose a channel interaction module (CIM), which combines an efficient channel attention network with global response normalization (GRN) in parallel to recalibrate features, enabling the network to learn crucial information at both the local and global scales. Additionally, we introduce a new silkworm ReID dataset for network training and evaluation. The experimental results demonstrate that WormNet achieves an mAP value of 54.8% and a rank-1 value of 91.4% on the dataset, surpassing both state-of-the-art and related networks. This study offers a valuable reference for ReID in insects and other organisms.