Research map / Detection and segmentation / Open-set and few-shot detection
CLIPSym: Delving into Symmetry Detection with CLIP
Tinghan Yang, Md Ashiqur Rahman, Raymond A. Yeh
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 CLIP pre-training instead of training from scratch (no-text) (ablation) | CLIPSym · DENDI reflection | F1: 32.1 -> 63.7 | Improved |
| add CLIP pre-training instead of ImageNet (no-text) (ablation) | CLIPSym no-text · DENDI reflection | F1: 54.8 -> 63.7 | Improved |
| add text encoder conditioning with grouped prompts (ablation) | CLIPSym · DENDI reflection | F1: 63.7 -> 66.5 | Improved |
| add text conditioning (ablation) | CLIPSym · DENDI rotation | F1: 17.7 -> 25.1 | Improved |
| add text conditioning (ablation) | CLIPSym · SDRW reflection | F1: 46.8 -> 51.8 | Improved |
| replace non-equivariant decoder with G-conv equivariant decoder (ablation) | CLIPSym · DENDI reflection | F1: 62.9 -> 66.5 | Improved |
| replace non-equivariant decoder with equivariant decoder (ablation) | CLIPSym · DENDI reflection | Consistency: 0.093 -> 0.082 | Improved |
| equivariant decoder vs non-equivariant decoder with 8x channels (ablation) | CLIPSym · DENDI reflection | F1: 64.9 -> 66.5 | Improved |
| use 25 prompts of 4 tokens instead of single phrase 'reflection axis' (ablation) | CLIPSym · DENDI reflection | F1: 64.4 -> 66.5 | Improved |
| increase to 50 prompts with 16 tokens (ablation) | CLIPSym · DENDI reflection | F1: 66.5 -> 64.4 | Hurt |
| make image encoder trainable (text frozen) (ablation) | CLIPSym · DENDI reflection | F1: 59.4 -> 65.3 | Improved |
| make text encoder trainable in addition to image encoder (ablation) | CLIPSym · DENDI reflection | F1: 65.3 -> 66.5 | Improved |
Closest papers by meaning
- R2Det: Exploring Relaxed Rotation Equivariance in 2D Object Detection (ICLR 2025 · similarity 0.86)
- Demystifying CLIP Data (ICLR 2024 · similarity 0.85)
- Calibrating Multi-modal Representations: A Pursuit of Group Robustness without Annotations (CVPR 2024 · similarity 0.84)
- SmartCLIP: Modular Vision-language Alignment with Identification Guarantees (CVPR 2025 · similarity 0.84)
- Just a Few Glances: Open-Set Visual Perception with Image Prompt Paradigm (AAAI 2025 · similarity 0.84)
- CLIPSelf: Vision Transformer Distills Itself for Open-Vocabulary Dense Prediction (ICLR 2024 · similarity 0.84)
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.
