Research map / Vision-language and multimodal / CLIP and zero-shot
OpenKD: Opening Prompt Diversity for Zero- and Few-shot Keypoint Detection
Changsheng Lu, Zheyuan Liu, Piotr Koniusz
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 auxiliary interpolated keypoints to visual training (ablation) | OpenKD · Animal Pose 1-shot novel | PCK@0.1: 21.36 -> 47.54 | Improved |
| add LLM-interpolated auxiliary texts to text training (ablation) | OpenKD · Animal Pose 0-shot novel | PCK@0.1: 25.60 -> 63.14 | Improved |
| use chain-of-thought prompting for LLM text interpolation (ablation) | OpenKD · AwA 0-shot novel | PCK@0.1: 73.80 -> 78.30 | Improved |
| select interpolated text by visual-text correlation instead of false text control (ablation) | OpenKD · AwA 0-shot novel | PCK@0.1: 78.30 -> 62.12 | Hurt |
| add false text control sampling (FTC) rather than correlation selection (ablation) | OpenKD · AwA 0-shot novel | PCK@0.1: 62.12 -> 78.64 | Improved |
| repeat LLM reasoning R=3 instead of R=1 (ablation) | OpenKD · AwA 0-shot novel | PCK@0.1: 67.34 -> 78.64 | Improved |
| increase R from 3 to 10 (ablation) | OpenKD · AwA 0-shot novel | PCK@0.1: 78.64 -> 76.61 | No clear effect |
| add intra-text contrastive loss L_tt (ablation) | OpenKD · AwA 0-shot novel | PCK@0.1: 76.71 -> 78.28 | Improved |
| add visual-text alignment L_vt without stop-gradient (ablation) | OpenKD · AwA 0-shot novel | PCK@0.1: 78.28 -> 77.10 | Hurt |
| add stop gradient on text features in L_vt (ablation) | OpenKD · AwA 0-shot novel | PCK@0.1: 77.10 -> 78.64 | Improved |
| add visual-visual contrastive loss L_vv (ablation) | OpenKD · AwA | PCK@0.1: 78.64 -> 78.60 | No clear effect |
| add LLM parsing of diverse text prompts (ablation) | OpenKD · Animal Pose 0-shot novel, diverse text | PCK@0.1: 13.89 -> 61.45 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- CLAMP — CLAMP: Prompt-Based Contrastive Learning for Connecting Language and Animal Pose (CVPR 2023) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- KptLLM: Unveiling the Power of Large Language Model for Keypoint Comprehension (NeurIPS 2024 · similarity 0.89)
- GKDT: General Keypoint Detection Transformer (ECCV 2026 · similarity 0.88)
- Just a Few Glances: Open-Set Visual Perception with Image Prompt Paradigm (AAAI 2025 · similarity 0.87)
- Detect Any Keypoints: An Efficient Light-Weight Few-Shot Keypoint Detector (AAAI 2024 · similarity 0.86)
- Visual Diversity and Region-aware Prompt Learning for Zero-shot HOI Detection (NeurIPS 2025 · similarity 0.86)
- PromptKD: Unsupervised Prompt Distillation for Vision-Language Models (CVPR 2024 · similarity 0.85)
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.
