Research map / Detection and segmentation / Open-set and few-shot detection
DesCo: Learning Object Recognition with Rich Language Descriptions
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 |
|---|---|---|---|
| append GPT-3 descriptions to class names at inference only | GLIP-T · LVIS rare categories | APr: 20.8 -> 12.2 | Hurt |
| fine-tune with descriptions plus entity name (no context-sensitive queries) (ablation) | GLIP-T · LVIS MiniVal | APr: 20.8 -> 20.5 | No clear effect |
| drop entity name from description queries (ablation) | GLIP-T + descriptions · LVIS MiniVal | APr: 20.5 -> 25.6 | Helped |
| drop entity name from description queries (ablation) | GLIP-T + descriptions · Context sensitivity test | AP: 5.7 -> 10.7 | Helped |
| add LLM-generated hard negative captions (ablation) | GLIP-T + descriptions w/o name · LVIS MiniVal | APr: 25.6 -> 26.5 | Helped |
| add LLM-generated hard negative captions (ablation) | GLIP-T + descriptions w/o name · OmniLabel COCO | APd: 16.0 -> 16.6 | Helped |
| add hard negative captions (ablation) | GLIP-T + descriptions w/o name · Context sensitivity test | AP: 10.7 -> 10.5 | No clear effect |
| use larger GPT model to generate descriptions (ada to davinci) (ablation) | DesCo-GLIP · LVIS MiniVal | APr: 19.9 -> 30.8 | Helped |
| use GPT-curie instead of davinci descriptions (ablation) | DesCo-GLIP · LVIS MiniVal | AP: 34.6 -> 32.8 | Hurt |
| DesCo training vs GLIP-T | DesCo-GLIP · LVIS MiniVal | APr: 20.8 -> 30.8 | Helped |
| DesCo training vs FIBER-B | DesCo-FIBER · LVIS MiniVal | AP: 33.8 -> 39.5 | Helped |
| DesCo training vs GLIP-T | DesCo-GLIP · OmniLabel | APd: 16.4 -> 21.0 | Helped |
Closest papers by meaning
- Roboflow100-VL: A Multi-Domain Object Detection Benchmark for Vision-Language Models (NeurIPS 2025 · similarity 0.91)
- Generative Region-Language Pretraining for Open-Ended Object Detection (CVPR 2024 · similarity 0.90)
- Revisiting Few-Shot Object Detection with Vision-Language Models (NeurIPS 2024 · similarity 0.89)
- How to Evaluate the Generalization of Detection? A Benchmark for Comprehensive Open-Vocabulary Detection (AAAI 2024 · similarity 0.89)
- The Devil is in the Fine-Grained Details: Evaluating Open-Vocabulary Object Detectors for Fine-Grained Understanding (CVPR 2024 · similarity 0.89)
- Synthesize Diagnose and Optimize: Towards Fine-Grained Vision-Language Understanding (CVPR 2024 · similarity 0.89)
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.