I am trying to train a custom Yolov5 model with the goal of integrating it in the voxl-tflite-server. I am using the a pretrained model in different formats (Yolov5s, Yolov5m or Yolov5l) The model is quantized as explained in https://docs.modalai.com/voxl-tflite-server/ , after replacing the model file in the server, it works but the output detects random boxes in the input image.
It is a Voxl 2 I do not know if there are a new documentation regarding this topic. In the link above the details for Voxl 2 is using the Tensorflow 2.8.0.
Is there any way to train the same Yolov5 model that is implemented in the voxl-tflite-server to avoid further coding in the server?
Thanks in advance