Vision Transformers with Mixed-Resolution Tokenization
This work addresses computational inefficiency in vision models for researchers and practitioners, though it is incremental as it builds on existing Vision Transformer architectures.
The authors tackled the problem of inefficient tokenization in Vision Transformers by introducing a mixed-resolution tokenization scheme that processes low-saliency areas at lower resolution, achieving substantial accuracy gains on image classification under controlled computational budgets.
Vision Transformer models process input images by dividing them into a spatially regular grid of equal-size patches. Conversely, Transformers were originally introduced over natural language sequences, where each token represents a subword - a chunk of raw data of arbitrary size. In this work, we apply this approach to Vision Transformers by introducing a novel image tokenization scheme, replacing the standard uniform grid with a mixed-resolution sequence of tokens, where each token represents a patch of arbitrary size. Using the Quadtree algorithm and a novel saliency scorer, we construct a patch mosaic where low-saliency areas of the image are processed in low resolution, routing more of the model's capacity to important image regions. Using the same architecture as vanilla ViTs, our Quadformer models achieve substantial accuracy gains on image classification when controlling for the computational budget. Code and models are publicly available at https://github.com/TomerRonen34/mixed-resolution-vit .