Good Token Hunting: A Hitchhiker's Guide to Token Selection for Visual Geometry Transformers
2026-05-22 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionArtificial IntelligenceGraphicsMachine LearningRobotics
AI summaryⓘ
The authors study a type of AI model called visual geometry transformers used for building 3D scenes from many images. These models get slower as they process more images because they look at all parts of the input at once. To fix this, the authors propose a method that smartly picks only the most important image parts to focus on, in two steps: choosing whole images first, then picking key details within those images. Their approach speeds up the model a lot without losing accuracy, showing a way to make these 3D reconstructions more efficient.
visual geometry transformersmulti-view 3D reconstructionglobal attentiontoken selectioninter-frame selectionintra-frame selectionentropylayer-aware sparsificationspeed-accuracy trade-offfeed-forward networks
Authors
Shuhong Zheng, Michael Oechsle, Erik Sandström, Marie-Julie Rakotosaona, Federico Tombari, Igor Gilitschenski
Abstract
Visual geometry transformers have become powerful architectures for multi-view 3D reconstruction, enabling joint prediction of multiple 3D attributes in a feed-forward manner. However, their computational cost grows quadratically with the input sequence length due to the global attention layers inside these models. This limits both their scalability and efficiency. In this work, we address this challenge with a simple yet general strategy: restricting the number of key/value tokens that each query interacts with during global attention. To achieve effective token selection, we introduce a two-stage framework. First, an inter-frame selection step operates at the frame level to identify frames that should be preserved. Second, an intra-frame selection step further discards more redundant tokens within the selected frames. Our analysis highlights the advantage of a diversity-based strategy for inter-frame selection, which ensures broad coverage of the scene. For intra-frame selection, we show that layer-aware sparsification is necessary, with the selection process guided by the entropy of the global attention pattern. Our approach offers a superior speed-accuracy trade-off compared to existing solutions. Extensive experiments show that it accelerates visual geometry transformers by over 85% for scenes with 500 images while maintaining, or even improving, baseline performance, which hints that how our token selection strategy can play a crucial role in future applications of visual geometry transformers. Our project website is available at https://zsh2000.github.io/good-token-hunting.github.io.