Vision

Computer vision methods that are useful background for modern multimodal systems.

Research category

Computer vision methods that are useful background for modern multimodal systems.

3Papers

9Resource links

2026.09Latest month

1 paper

Object Detection

2022.03 Object Detection

DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object Detection

DINO improves DETR-like object detectors by introducing contrastive denoising training, mixed query selection for anchor initialization, and a look-forward-twice box prediction scheme, achieving state-of-the-art results on COCO with significantly reduced model and data requirements.

Paper Code

1 paper

OCR

2026.09 OCR

All-in-One Multilingual Scene Text Recognition with Script-aware Mixture-of-Experts

This paper introduces ScriptMoE, a multilingual scene text recognizer that shares a visual encoder and routes each image to two script-aligned decoder experts alongside a shared expert. It also builds TextMuSS-10M, a synthetic dataset covering 10 scripts and 229 languages, to address scarce training data and improve multilingual recognition without the cost of large vision-language models.

Paper Code Hugging Face

1 paper

Semantic Correspondence

2026.05 Semantic Correspondence

SOCO: Benchmarking Semantic Object Correspondence in Vision Foundation Models

SOCO introduces a taxonomy-driven benchmark with consistent, functionally meaningful keypoint annotations across 100 categories and more than one million correspondence pairs. Its evaluation reveals gaps in cross-category transfer and object-part geometry understanding while showing that correspondence performance strongly predicts dense downstream task performance.

Paper Project Code Hugging Face