Patch Slimming for Efficient Vision Transformers
Citations Over TimeTop 10% of 2022 papers
Abstract
This paper studies the efficiency problem for visual transformers by excavating redundant calculation in given networks. The recent transformer architecture has demonstrated its effectiveness for achieving excellent performance on a series of computer vision tasks. However, similar to that of convolutional neural networks, the huge computational cost of vision transformers is still a severe issue. Considering that the attention mechanism aggregates different patches layer-by-layer, we present a novel patch slimming approach that discards useless patches in a topdown paradigm. We first identify the effective patches in the last layer and then use them to guide the patch selection process of previous layers. For each layer, the impact of a patch on the final output feature is approximated and patches with less impacts will be removed. Experimental results on benchmark datasets demonstrate that the proposed method can significantly reduce the computational costs of vision transformers without affecting their performances. For example, over 45% FLOPs of the ViT-Ti model can be reduced with only 0.2% top-1 accuracy drop on the ImageNet dataset.
Related Papers
- → FLOPs-efficient filter pruning via transfer scale for neural network acceleration(2021)12 cited
- → Towards Flops-constrained Face Recognition(2019)
- → An Interdisciplinary Study on the Correlation Between FLOPS Performance and VFX CG/CGI Workflow Improvements: FLOPS Improvements in the Early 90s and Examples of CG/CGI VFX(2022)
- → Chapter 6: Flop counts(2022)