# Win-Win: Training High-Resolution Vision Transformers from Two Windows

arXiv: https://arxiv.org/abs/2310.00632 · HTML: https://mlsanity.com/paper/2310.00632/

## Key facts

- Cited by 12 papers in the map (data through 14 Oct 2026).
- Builds on 67 of its 74 references that are in the map.
- First posted on arXiv 1 Oct 2023.

## Builds on (most cited first, 30 of 67)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [Swin Transformer: Hierarchical Vision Transformer using Shifted Windows](https://mlsanity.com/paper/2103.14030/) (2021)
- [Segment Anything](https://mlsanity.com/paper/2304.02643/) (2023)
- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [Masked Autoencoders Are Scalable Vision Learners](https://mlsanity.com/paper/2111.06377/) (2021)
- [Emerging Properties in Self-Supervised Vision Transformers](https://mlsanity.com/paper/2104.14294/) (2021)
- [RoFormer: Enhanced Transformer with Rotary Position Embedding](https://mlsanity.com/paper/2104.09864/) (2021)
- [Training data-efficient image transformers & distillation through attention](https://mlsanity.com/paper/2012.12877/) (2020)
- [A ConvNet for the 2020s](https://mlsanity.com/paper/2201.03545/) (2022)
- [RAFT: Recurrent All-Pairs Field Transforms for Optical Flow](https://mlsanity.com/paper/2003.12039/) (2020)
- [Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions](https://mlsanity.com/paper/2102.12122/) (2021)
- [Vision Transformers for Dense Prediction](https://mlsanity.com/paper/2103.13413/) (2021)
- [Rethinking Attention with Performers](https://mlsanity.com/paper/2009.14794/) (2020)
- [Linformer: Self-Attention with Linear Complexity](https://mlsanity.com/paper/2006.04768/) (2020)
- [Rethinking Semantic Segmentation from a Sequence-to-Sequence Perspective with Transformers](https://mlsanity.com/paper/2012.15840/) (2020)
- [Multi-Task Learning Using Uncertainty to Weigh Losses for Scene Geometry and Semantics](https://mlsanity.com/paper/1705.07115/) (2017)
- [VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training](https://mlsanity.com/paper/2203.12602/) (2022)
- [BDD100K: A Diverse Driving Dataset for Heterogeneous Multitask Learning](https://mlsanity.com/paper/1805.04687/) (2018)
- [Swin Transformer V2: Scaling Up Capacity and Resolution](https://mlsanity.com/paper/2111.09883/) (2021)
- [SimMIM: A Simple Framework for Masked Image Modeling](https://mlsanity.com/paper/2111.09886/) (2021)
- [Tokens-to-Token ViT: Training Vision Transformers from Scratch on ImageNet](https://mlsanity.com/paper/2101.11986/) (2021)
- [FlowNet: Learning Optical Flow with Convolutional Networks](https://mlsanity.com/paper/1504.06852/) (2015)
- [Token Merging: Your ViT But Faster](https://mlsanity.com/paper/2210.09461/) (2022)
- [CvT: Introducing Convolutions to Vision Transformers](https://mlsanity.com/paper/2103.15808/) (2021)
- [Multiscale Vision Transformers](https://mlsanity.com/paper/2104.11227/) (2021)
- [A Large Dataset to Train Convolutional Networks for Disparity, Optical Flow, and Scene Flow Estimation](https://mlsanity.com/paper/1512.02134/) (2015)
- [DynamicViT: Efficient Vision Transformers with Dynamic Token Sparsification](https://mlsanity.com/paper/2106.02034/) (2021)
- [Exploring Plain Vision Transformer Backbones for Object Detection](https://mlsanity.com/paper/2203.16527/) (2022)
- [CCNet: Criss-Cross Attention for Semantic Segmentation](https://mlsanity.com/paper/1811.11721/) (2018)

## Built on by (most cited first, 12 of 12)

- [SEA-RAFT: Simple, Efficient, Accurate RAFT for Optical Flow](https://mlsanity.com/paper/2405.14793/) (2024)
- [Pow3R: Empowering Unconstrained 3D Reconstruction with Camera and Scene Priors](https://mlsanity.com/paper/2503.17316/) (2025)
- [WAFT: Warping-Alone Field Transforms for Optical Flow](https://mlsanity.com/paper/2506.21526/) (2025)
- [FlowSeek: Optical Flow Made Easier with Depth Foundation Models and Motion Bases](https://mlsanity.com/paper/2509.05297/) (2025)
- [Thicker and Quicker: A Jumbo Token for Fast Plain Vision Transformers](https://mlsanity.com/paper/2502.15021/) (2025)
- [Tile-Based ViT Inference with Visual-Cluster Priors for Zero-Shot Multi-Species Plant Identification](https://mlsanity.com/paper/2507.06093/) (2025)
- [HRSAM: Efficient Interactive Segmentation in High-Resolution Images](https://mlsanity.com/paper/2407.02109/) (2024)
- [Do VLMs Need Vision Transformers? Evaluating State Space Models as Vision Encoders](https://mlsanity.com/paper/2603.19209/) (2026)
- [FlowIt: Global Matching via Hierarchical Transformers and Optimal Transport for Optical Flow](https://mlsanity.com/paper/2603.28759/) (2026)
- [ELSA: Exact Linear-Scan Attention for Fast and Memory-Light Vision Transformers](https://mlsanity.com/paper/2604.23798/) (2026)
- [HRDiT: Training-Free High-Resolution Image Generation with Off-the-Shelf Diffusion Transformer Models](https://mlsanity.com/paper/2608.07003/) (2026)
- [FreeFlow: A Bias-free Hierarchical Transformer for Optical Flow Estimation](https://mlsanity.com/paper/2609.11486/) (2026)

## Abstract (arXiv)

Transformers have become the standard in state-of-the-art vision architectures, achieving impressive performance on both image-level and dense pixelwise tasks. However, training vision transformers for high-resolution pixelwise tasks has a prohibitive cost. Typical solutions boil down to hierarchical architectures, fast and approximate attention, or training on low-resolution crops. This latter solution does not constrain architectural choices, but it leads to a clear performance drop when testing at resolutions significantly higher than that used for training, thus requiring ad-hoc and slow post-processing schemes. In this paper, we propose a novel strategy for efficient training and inference of high-resolution vision transformers. The key principle is to mask out most of the high-resolution inputs during training, keeping only N random windows. This allows the model to learn local interactions between tokens inside each window, and global interactions between tokens from different windows. As a result, the model can directly process the high-resolution input at test time without any special trick. We show that this strategy is effective when using relative positional embedding such as rotary embeddings. It is 4 times faster to train than a full-resolution network, and it is straightforward to use at test time compared to existing approaches. We apply this strategy to three dense prediction tasks with high-resolution data. First, we show on the task of semantic segmentation that a simple setting with 2 windows performs best, hence the name of our method: Win-Win. Second, we confirm this result on the task of monocular depth prediction. Third, we further extend it to the binocular task of optical flow, reaching state-of-the-art performance on the Spring benchmark that contains Full-HD images with an order of magnitude faster inference than the best competitor.