Hi fast.ai forum users,
I would like to use the VGG16 Places365 (https://github.com/CSAILVision/places365) in Keras, but I am having a hard time figure out how to convert the provided Caffe model to a Keras model. Has anybody already done this or would be able to point me in the right direction on how to do this?
Thank you in advance for your help!
Do you just want to convert architecture? Or do you want to use pretrained weights from Caffe? If former, then it shouldn’t be too hard, but if latter, I guess it is not possible to use pretrained weights from Caffe in Keras, so you will need to train your model from scratch.
Yes I do want to use the pretrained weights from Caffe
If you port an architecture from one framework to another, you need to retrain the whole model on the whole data set.
Why can’t you go with Caffe with this particular model?