Knowledge Distillation in YOLOX-ViT for Side-Scan Sonar Object Detection

In this paper we present YOLOX-ViT, a novel object detection model, andinvestigate the efficacy of knowledge distillation for model size reductionwithout sacrificing performance. Focused on underwater robotics, our researchaddresses key questions about the viability of smaller models and the impact ofthe visual transformer layer in YOLOX. Furthermore, we introduce a newside-scan sonar image dataset, and use it to evaluate our object detector'sperformance. Results show that knowledge distillation effectively reduces falsepositives in wall detection. Additionally, the introduced visual transformerlayer significantly improves object detection accuracy in the underwaterenvironment. The source code of the knowledge distillation in the YOLOX-ViT isat https://github.com/remaro-network/KD-YOLOX-ViT.