Thread-Efficient Decoding for Neural Texture Compression
Computer Vision and Pattern RecognitionGraphics
Summary
The authors worked on a way to compress textures in graphics that is both smaller in size and faster to use on GPUs. They created a shared decoder system that groups similar textures together to reduce inefficiencies during rendering. To make this system work well, they trained it gradually and used smart clustering based on image features. Their method made rendering up to 8.5 times faster on a specific GPU, while keeping the image quality good. Overall, their approach balances speed and quality better than previous methods.
Authors
Janarbek Matai, Sho Ikeda, Lukasz Lipski, Takahiro Harada
Abstract
Neural texture compression (NTC) achieves higher compression ratios than BCn formats but suffers from GPU thread divergence, which significantly reduces runtime performance. In this work, we propose a shared decoder MLP architecture -- trained with a gradual decoder freezing schedule -- combined with texture clustering to reduce thread divergence by 25%-52% while preserving rendering quality. We evaluate our method on over 500 textures and multiple real rendering scenes, demonstrating up to 8.48x speedup on the Radeon RX 9070 XT GPU compared to non-shared baselines. Our key contributions include: (1) a unified shared decoder architecture that reduces divergence by grouping textures; (2) a training recipe with gradual decoder freezing that improves stability and reconstruction accuracy; (3) a semantic clustering strategy using CLIP embeddings that groups similar textures for effective decoder sharing; and (4) comprehensive performance and ablation studies validating our approach.