I trained a YOLO v5 object detection model on my custom dataset, then converted it to TensorFlow using the export function and to TensorFlow lite using tf.convertor.
I want to use the TFLite model in an android app, however, the problem is the file size is relatively large (~ 27.5 MB).
Is there a way to make its size smaller without affecting the accuracy?
What I have tried:
I read about quantization and pruning but I don't know what would be best to not affect my model's performance.