Soft then Hard: Rethinking the Quantization in Neural Image Compression

               Zongyu Guo 1 Zhizheng Zhang 1 Runsen Feng 1 Zhibo Chen 1

             Abstract                Quantization is one of the key challenges for neural image
                                 compression. Since the gradient of quantization is zero al-
   Quantization is one of the core components in        most everywhere, it makes the standard back-propagation
   lossy image compression. For neural image com-     ...