I just wanted to share this new really cool paper on a new ConvNet that does very well. Just look at this graph!
Here’s the paper: https://arxiv.org/abs/1905.11946
Here’s the blog post: https://ai.googleblog.com/2019/05/efficientnet-improving-accuracy-and.html
And here’s the official implementation: https://github.com/tensorflow/tpu/tree/master/models/official/efficientnet
Seems like a big step forward!
Quoc Le’s tweet announcing it sounds like a click bait advert for deep learning nerds