Jason Antic

1/ Higher Top 1 accuracy in image classification actually isn't a great indicator that a new vision model is going to work out well in practice in an image to image task (like DeOldify). I've learned this the hard way after getting excited about a new shiny model many times!

2/ EfficientNets come to mind- I just have not been seeing the benefits relative to the classification performance. But this really applies generally, even within a set of architectures that have been trained the same way (FaceBook's wsl models, BiTM, etc).

3/ In those cases, you just don't really know until you actually try. You may be surprised, I can tell you that much! So don't just pick the one with the highest Top-1 accuracy in ImageNet...

Any better metrics that you might not have come up with along the way? Or sets of metrics?

2/ ... learning tasks. Specifically- the larger and more varied the training, the better. This is basically the Facebook WSL models' secret ingredient too. I think everything else is basically noise that you have to sift through with experiments.