Luke Melas-Kyriazi

icon indicating a website link https://lukemelas.github.io/

icon company Anysphere (Cursor) icon location San Francisco, CA icon location Building @getcursor. Previously Rhodes Scholar and a PhD student at Oxford University. Always open to meeting interesting people!

Results 60 comments of Luke Melas-Kyriazi

what is meaning of 'gts, res' in code?

Ground truths, result

TPU Support

Cool, thanks for the response!

TPU Support

I'll leave it open as an enhancement, but feel free to close it at any time if you would like.

AutoEncoder using the EfficientNet

Thanks for this PR! Very interesting. I'll have to think about whether this should be integrated into the main repo or whether it should be a standalone repo. Either way,...

Feature vector for image similarity measure

That feature vector still has spacial information, so you might try averaging over the spacial dimensions (e.g. use `torch.mean` to make the 1792x7x7-dim vector to a 1792-dim vector). Let me...

Feature vector for image similarity measure

Great! Keep me updated. On Mon, Aug 26, 2019, 8:53 PM drjtan wrote: > yes, averaging works. However, I am not sure if the feature vectors > obtained this way...

Normalize input using numpy only for onnx inference.

Hi, I'm happy to help. Can you give a complete reproducible example?

the inference speed is much slower than original TensorFlow code

Hello, Yes, this is expected because grouped convolutions in PyTorch are slow (the core devs are working on making them faster). See pytorch/pytorch#18631.

the inference speed is much slower than original TensorFlow code

Now with #44 you can export to ONNX. That may help in terms of inference speed, as it actually compiles a graph.

Can not load from pretrained?

What version of `efficientnet_pytorch` are you using?