-
[pdf]
[bibtex]@InProceedings{Tao_2023_ICCV, author = {Tao, Lvfang and Gao, Wei and Li, Ge and Zhang, Chenhao}, title = {AdaNIC: Towards Practical Neural Image Compression via Dynamic Transform Routing}, booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)}, month = {October}, year = {2023}, pages = {16879-16888} }
AdaNIC: Towards Practical Neural Image Compression via Dynamic Transform Routing
Abstract
Compressive autoencoders (CAEs) play an important role in deep learning-based image compression, but large-scale CAEs are computationally expensive. We propose a framework with three techniques to enable efficient CAE-based image coding: 1) Spatially-adaptive convolution and normalization operators enable block-wise nonlinear transform to spend FLOPs unevenly across the image to be compressed, according to a transform capacity map. 2) Just-unpenalized model capacity (JUMC) optimizes the transform capacity of each CAE block via rate-distortion-complexity optimization, finding the optimal capacity for the source image content. 3) A lightweight routing agent model predicts the transform capacity map for the CAEs by approximating JUMC targets. By activating the best-sized sub-CAE inside the slimmable supernet, our approach achieves up to 40% computational speed-up with minimal BD-Rate increase, validating its ability to save computational resources in a content-aware manner.
Related Material