Faster R-CNN-based diatom image recognition model in forensic identification

Aiming at the current problems of low accuracy and low efficiency of diatom recognition in forensic science, this project intends to construct a diatom image recognition model from four levels, namely, dataset construction, model design, training and testing, and carry out experimental research on the model to explore the model's performance and optimization methods in order to test its application value in judicial practice. The experiment proves that the method proposed in this paper can classify diatoms and extract the features of diatoms, which provides intelligent technical support for judicial appraisal work.