naver-ai/rope-vit
A research implementation applying rotary position embeddings (RoPE) to vision transformers for improved image classification and detection performance.
Not currently ranked — collecting fresh signals.
star history
This repository provides the official PyTorch implementation of RoPE-ViT, a research paper from NAVER AI Lab published at ECCV 2024. The work applies Rotary Position Embedding (RoPE), originally successful in language models for length extrapolation, to Vision Transformers. The implementation enables improved resolution extrapolation at inference time while maintaining accuracy, demonstrating gains across multiple computer vision benchmarks including ImageNet-1k classification, COCO object detection, and ADE-20k semantic segmentation.
Frequently asked
- What is naver-ai/rope-vit?
- A research implementation applying rotary position embeddings (RoPE) to vision transformers for improved image classification and detection performance.
- Is rope-vit open source?
- Yes — naver-ai/rope-vit is an open-source project tracked on heatdrop.
- What language is rope-vit written in?
- naver-ai/rope-vit is primarily written in Python.
- How popular is rope-vit?
- naver-ai/rope-vit has 467 stars on GitHub.
- Where can I find rope-vit?
- naver-ai/rope-vit is on GitHub at https://github.com/naver-ai/rope-vit.