Research map / Detection and segmentation / Open-set and few-shot detection
Grounded Language-Image Pre-Training
Check 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 language-aware deep fusion to the grounding reformulation (ablation) | GLIP-T on COCO · COCO 2017val, zero-shot | AP: 42.9 -> 44.9 | Helped |
| Add GoldG human-annotated grounding data to O365 pre-training (ablation) | GLIP-T · COCO 2017val, zero-shot | AP: 44.9 -> 46.7 | Helped |
| Add Cap4M web image-text data to O365 and GoldG pre-training (ablation) | GLIP-T · COCO 2017val, zero-shot | AP: 46.7 -> 46.3 | Hurt |
| Add GoldG grounding data to O365 pre-training (ablation) | GLIP-T · LVIS MiniVal | APr: 12.8 -> 19.5 | Helped |
| Add image-text data to O365 and GoldG pre-training (ablation) | GLIP-T · LVIS MiniVal | APr: 19.5 -> 21.4 | Helped |
| Add detection data to grounding pre-training (ablation) | GLIP-T on Flickr30K Entities · Flickr30K Entities test | Recall@1: 84.4 -> 85.5 | Helped |
| Add Cap4M image-text data to O365 and GoldG pre-training (ablation) | GLIP-T on Flickr30K Entities · Flickr30K Entities test | Recall@1: 85.5 -> 85.7 | Helped |
| Scale detection-only pre-training across four detection datasets instead of using O365 with GoldG (ablation) | GLIP-T · COCO 2017val | AP: 46.7 -> 46.3 | Hurt |
| Add GoldG to detection-data pre-training (ablation) | GLIP-T with VG without COCO · COCO 2017val | AP: 26.9 -> 29.2 | Helped |
| Add GoldG to detection-data pre-training (ablation) | GLIP-T with OpenImages · COCO 2017val | AP: 29.9 -> 33.6 | Helped |
| Use expressive prompt attributes describing the stingray as flat and round (ablation) | Zero-shot GLIP on Aquarium · Aquarium stingray | AP50: 4.6 -> 9.7 | Helped |
| Tune task-specific prompt embeddings rather than the full model (ablation) | GLIP-T and GLIP-L on ODinW · 13 ODinW datasets | average AP | No clear effect |
Closest papers by meaning
- GroundVLP: Harnessing Zero-Shot Visual Grounding from Vision-Language Pre-training and Open-Vocabulary Object Detection (AAAI 2024 · similarity 0.90)
- DoGA: Enhancing Grounded Object Detection via Grouped Pre-Training with Attributes (AAAI 2025 · similarity 0.88)
- Connecting the Dots: Training-Free Visual Grounding via Agentic Reasoning (AAAI 2026 · similarity 0.87)
- GeM-VG: Towards Generalized Multi-image Visual Grounding with Multimodal Large Language Models (AAAI 2026 · similarity 0.87)
- Roboflow100-VL: A Multi-Domain Object Detection Benchmark for Vision-Language Models (NeurIPS 2025 · similarity 0.87)
- LaViP: Language-Grounded Visual Prompting (AAAI 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.