3D Posture Estimation and Automatic Adjustment of Zebrafish Larvae

Due to the irregular surface of zebrafish larvae and the difficulty of fixing the body, there are high requirements for its injection target point and posture control. By establishing a zebrafish data set containing key points and three-dimensional posture annotations, a network model for zebrafish key point recognition is designed using convolutional neural networks. The posture angle information is further calculated through three-dimensional coordinate transformation. Experiments demonstrate that this method can obtain a more accurate zebrafish posture angle under the condition of posture change. In response to this method, an operating platform that can fix zebrafish juveniles and precisely adjust their posture is designed to lay the foundation for subsequent experimental operations.

Automated summary

The irregular surface of zebrafish larvae presents challenges for fixing their bodies, leading to high requirements for injection target points and posture control. A zebrafish dataset containing key points and posture annotations was established, and a neural network model designed for zebrafish key point recognition. Through three-dimensional coordinate transformation, posture angle information was accurately calculated, demonstrating improved accuracy in zebrafish posture angle determination under changing postures. An operating platform was also designed to fix zebrafish juveniles and precisely adjust their posture, laying the foundation for further experimental operations.