{"paper":{"arxiv_id":"2111.09883","title":"Swin Transformer V2: Scaling Up Capacity and Resolution","abstract":"Large-scale NLP models have been shown to significantly improve the performance on language tasks with no signs of saturation. They also demonstrate amazing few-shot capabilities like that of human beings. This paper aims to explore large-scale models in computer vision. We tackle three major issues in training and application of large vision models, including training instability, resolution gaps between pre-training and fine-tuning, and hunger on labelled data. Three main techniques are proposed: 1) a residual-post-norm method combined with cosine attention to improve training stability; 2) A log-spaced continuous position bias method to effectively transfer models pre-trained using low-resolution images to downstream tasks with high-resolution inputs; 3) A self-supervised pre-training method, SimMIM, to reduce the needs of vast labeled images. Through these techniques, this paper successfully trained a 3 billion-parameter Swin Transformer V2 model, which is the largest dense vision model to date, and makes it capable of training with images of up to 1,536x1,536 resolution. It set new performance records on 4 representative vision tasks, including ImageNet-V2 image classification, COCO object detection, ADE20K semantic segmentation, and Kinetics-400 video action classification.","primary_category":"cs.CV","venue":"CVPR 2022","published_at":null,"latest_version":1,"withdrawn":false},"latest_version":{"id":"4386006f-3bfe-4052-a8e0-c8c940ea8be2","version":1,"source_url":"https://arxiv.org/abs/2111.09883","rendered_html_url":null,"rendering_engine":null},"verdict":{"id":"23692ad6-87c2-4991-9664-8e9172473d72","kind":"POST","status":"partial","score":0.9076305220883535,"confidence":0.55,"agent_version":"v0.1.0-swinv2-imagenet-microslice","computed_at":"2026-05-14T23:48:47.059Z","is_current":true,"claim_citation":{"paper_arxiv_id":"2111.09883","section":"Table 5","row":"SwinV2-T (window 8, 256²)","column":"ImageNet-1k val top-1","reported_value":81.8,"reported_metric":"accuracy","quoted_text":"81.8","pdf_page":8,"notes":"Table 5 of arXiv:2111.09883 reports SwinV2-T (256², window 8) ImageNet-1k val top-1 = 81.8. Driver evaluates the official `microsoft/swinv2-tiny-patch4-window8-256` checkpoint on a 3-slice ImageNet-1k val micro-slice (Imagenette fallback on 401-gated environments). PROTOCOL_MATCH is `proxy` (dataset-size)."},"protocol_match":"proxy"},"verdicts":{"post":{"id":"23692ad6-87c2-4991-9664-8e9172473d72","kind":"POST","status":"partial","score":0.9076305220883535,"confidence":0.55,"agent_version":"v0.1.0-swinv2-imagenet-microslice","computed_at":"2026-05-14T23:48:47.059Z","is_current":true,"claim_citation":{"paper_arxiv_id":"2111.09883","section":"Table 5","row":"SwinV2-T (window 8, 256²)","column":"ImageNet-1k val top-1","reported_value":81.8,"reported_metric":"accuracy","quoted_text":"81.8","pdf_page":8,"notes":"Table 5 of arXiv:2111.09883 reports SwinV2-T (256², window 8) ImageNet-1k val top-1 = 81.8. Driver evaluates the official `microsoft/swinv2-tiny-patch4-window8-256` checkpoint on a 3-slice ImageNet-1k val micro-slice (Imagenette fallback on 401-gated environments). PROTOCOL_MATCH is `proxy` (dataset-size)."},"protocol_match":"proxy"},"pre":null}}