Research map / Vision-language and multimodal / CLIP and zero-shot
GLIPv2: Unifying Localization and Vision-Language Understanding
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add inter-image word-region contrastive loss (ablation) | GLIP pre-training with intra-image region-word contrastive loss · COCO, ODinW, and LVIS object detection | object detection performance | Helped |
| Add MLM loss (ablation) | GLIPv2 pre-training · Flickr30K, VQA, and image captioning | VL understanding performance | Helped |
| Add weak image-text pair data (Cap) (ablation) | GLIPv2 pre-training · COCO, Flickr30K, ODinW, and LVIS | task performance | Helped |
| Add inter-image region-word contrastive loss at fixed data scale (ablation) | GLIPv2 pre-training · Large-domain-gap tasks | representation performance | Helped |
| Use MLM-only pre-training instead of no pre-training (ablation) | Grounded captioning · COCO Caption | BLEU-4: 35.4 -> 33.4 | Hurt |
| Use localization and intra-/inter-image contrastive pre-training instead of no pre-training (ablation) | Grounded captioning · COCO Caption | BLEU-4: 35.4 -> 36.6 | Helped |
| Use MLM-only pre-training instead of no pre-training (ablation) | Grounded captioning · Flickr30K Grounding | R@1: 77.0 -> 70.9 | Hurt |
| Use localization and intra-/inter-image contrastive pre-training instead of no pre-training (ablation) | Grounded captioning · Flickr30K Grounding | R@1: 77.0 -> 80.8 | Helped |
Closest papers by meaning
- RegionGPT: Towards Region Understanding Vision Language Model (CVPR 2024 · similarity 0.88)
- Mechanisms of Object Localization in Vision-Language Models (CVPR 2026 · similarity 0.88)
- VladVA: Discriminative Fine-tuning of LVLMs (CVPR 2025 · similarity 0.88)
- Your Large Vision-Language Model Only Needs A Few Attention Heads For Visual Grounding (CVPR 2025 · similarity 0.88)
- Cyclic Contrastive Knowledge Transfer for Open-Vocabulary Object Detection (ICLR 2025 · similarity 0.88)
- VisionLLM v2: An End-to-End Generalist Multimodal Large Language Model for Hundreds of Vision-Language Tasks (NeurIPS 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.