Research map / Vision-language and multimodal / CLIP and zero-shot
CLAMP: Prompt-Based Contrastive Learning for Connecting Language and Animal Pose
Xu Zhang, Wen Wang, Zhe Chen, Yufei Xu, Jing Zhang, Dacheng Tao
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 |
|---|---|---|---|
| Use CLIP rather than ImageNet pre-training | SimpleBaseline with ResNet-50 · AP-10K supervised | AP: 70.2 -> 70.9 | Improved |
| Use CLAMP instead of the CLIP-pretrained SimpleBaseline | ResNet-50 animal pose estimation · AP-10K supervised | AP: 70.9 -> 72.9 | Improved |
| Use CLAMP instead of the CLIP-pretrained SimpleBaseline | ViT-Base animal pose estimation · AP-10K supervised | AP: 72.6 -> 74.3 | Improved |
| Use CLIP rather than ImageNet pre-training | SimpleBaseline with ResNet-50 · Animal-Pose supervised | AP: 68.7 -> 70.8 | Improved |
| Use CLAMP instead of the CLIP-pretrained SimpleBaseline | ResNet-50 animal pose estimation · Animal-Pose supervised | AP: 70.8 -> 72.5 | Improved |
| Use CLAMP instead of the CLIP-pretrained SimpleBaseline | ViT-Base animal pose estimation · Animal-Pose supervised | AP: 72.3 -> 74.3 | Improved |
| Use CLAMP instead of SimpleBaseline | ResNet-50 animal pose estimation · AP-10K, Bovidae train → Canidae test | AP: 41.3 -> 46.9 | Improved |
| Use CLAMP instead of SimpleBaseline | ResNet-50 animal pose estimation · AP-10K, Canidae train → Felidae test | AP: 39.6 -> 48.4 | Improved |
| Add spatial-level contrastive loss (ablation) | CLAMP with CLIP-pretrained ResNet-50 · AP-10K supervised | AP: 70.9 -> 72.1 | Improved |
| Add feature-level contrastive loss (ablation) | CLAMP with spatial-level loss · AP-10K supervised | AP: 72.1 -> 72.6 | Improved |
| Add prompt encoder (ablation) | CLAMP with spatial- and feature-level losses · AP-10K supervised | AP: 72.6 -> 72.9 | Improved |
| Replace prompt embedding with a trainable matrix (ablation) | CLAMP with ViT-Base · AP-10K supervised | AP: 74.3 -> 72.8 | Hurt |
Later papers that compared with this method
3 later accepted papers report CLAMP as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- Probabilistic Prompt Distribution Learning for Animal Pose Estimation (CVPR 2025 · better on 11 of 11 shared results)
- KptLLM: Unveiling the Power of Large Language Model for Keypoint Comprehension (NeurIPS 2024 · better on 2 of 2 shared results)
- OpenKD: Opening Prompt Diversity for Zero- and Few-shot Keypoint Detection (ECCV 2024 · 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
- Probabilistic Prompt Distribution Learning for Animal Pose Estimation (CVPR 2025 · similarity 0.83)
- CoPL: Contextual Prompt Learning for Vision-Language Understanding (AAAI 2024 · similarity 0.81)
- Contrastive Localized Language-Image Pre-Training (ICML 2025 · similarity 0.80)
- UniPose: A Unified Multimodal Framework for Human Pose Comprehension, Generation and Editing (CVPR 2025 · similarity 0.80)
- Concept-Guided Prompt Learning for Generalization in Vision-Language Models (AAAI 2024 · similarity 0.80)
- LaViP: Language-Grounded Visual Prompting (AAAI 2024 · similarity 0.80)
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.
