PRUNING

[NonlinearNimesh]

HI, I want to shrink the size of the trained east model with pruning, can anyone please tell me how I can do this. I tried quantization but after implementing the quantization size of the model is 30MB approx. I Want the size to be around 10MB.

Thanks

[francotheengineer]

TF Lite Quantization with TF optimization . This is a non-trivial task and you may see better results using a smaller backbone (ResNet50 is not small). Maybe MobileNet as a feature extractor - you'll have to figure the best layers to use as embeddings.