Philipp Luniak

I think it might be easier to re-train your with the ResNet50 implementation of this repo. Alternatively you could add the decoder to your existing model. It's not that difficult.

Are you sure it's not working? I'm using the repo with TF2.3.0.

Sorry, can't help you with that.

Thanks for the input @LingDong- . I had a look into it and I'm afraid that is out of scope for me at the moment. I'm more of a pure...

@Meight @bonlime Have you tried finetuning the whole model or just the last couple of layers? In the link posted by bonlime they say that the problem stems from the...

@Meight Many thanks for pointing this out. Surely you saved me a lot of time! May I ask which other models have evolved since them that performed equally well for...

@kritiyer @rdiazgar @Licini Thanks for your input! Can you please tell which versions of TF/Keras you were using? Philipp

I have also successfully fine-tuned this model. I did nothing special: TF.1.13.1-GPU, Keras 2.2.4, binary_cross_entropy, Adamax(default params), labels_shape(height,width,no_classes). Keras.ImageDataGenerator and class_weights also work. I passed in numpy arrays. Converges quickly...

@rdiazgar Thanks. Makes sense :-)

@jingxin328 Try downsizing the images to sth like 200*400 and use OS=16 instead. I'm on a GTX 1080 Ti (12 GB video ram) and I'd also run OOM with your...