Inspired by human vision's architecture (high-resolution foveal center, low-resolution periphery), we embedded this retinotopic transformation into CNN architectures, allowing to actively scan the image. This gives literally a new look to #ConvNets !
Quantization Variation: A New Perspective on Training Transformers with Low-Bit Precision
Xijie Huang, Zhiqiang Shen, Pingcheng Dong, Kwang-Ting Cheng
Action editor: Naigang Wang
https://openreview.net/forum?id=MHfoA0Qf6g
#quantization #convnets #imagenet
Join #CosmiQ once again to learn the fastest way to find boats in the open water! #SpaceNet (Hint: Not #ConvNets!)
medium.com/the-downlinq/object-dete...
Really cool work on using #ConvNets! Generate "photos" from sketches! The full paper: https://arxiv.org/pdf/1606.03073v1.pdf twitter.com/SriChandrasekar/status/7...
