Contamination MapVision / Model
UniVAD
- Organisation
- Institute of Automation, Chinese Academy of Sciences et al.
- Family
- UniVAD
- Released
- 2024-12-04 (arXiv v1 submission date (https://arxiv.org/abs/2412.03342))
- Weights
- Unknown
- Links
- PaperCode
Training data disclosure: Full list
Level determined for 0% of benchmark cells.
- TestTest data seen0
- SeenSeen, split unclear0
- TrainTraining data seen0
- Related data seen0
- NoNot seen0
- ?Unknown8
“In few-normal-shot setting, we do not conduct any further training on anomaly detection datasets for UniVAD.”
Built on
Everything these base models may have seen is inherited by UniVAD.
-
OpenAI CLIP
· disclosure: Partial list
Evidence
“utilize two widely used vision encoders, CLIP-L/14@336px and DINOv2-G/14, as our image encoders, with their parameters frozen.”
arxiv.org · paper · Section 4, Implementation Details · retrieved 2026-10-15 “pretrained = "openai"”
raw.githubusercontent.com · code · UniVAD.py · retrieved 2026-10-15 -
DINOv2
· disclosure: Partial list
Evidence
“utilize two widely used vision encoders, CLIP-L/14@336px and DINOv2-G/14, as our image encoders, with their parameters frozen.”
arxiv.org · paper · Section 4, Implementation Details · retrieved 2026-10-15 -
Segment Anything Model (SAM)
· disclosure: Partial list
Evidence
“wget https://huggingface.co/lkeab/hq-sam/resolve/main/sam_hq_vit_h.pth;”
raw.githubusercontent.com · readme · Prepare pretrained checkpoints · retrieved 2026-10-15
Other declared training data
-
Pre-training data of base model grounding-dino (not tracked here) (broad web data: may contain benchmark test sets) · split: unknown
Evidence
“wget https://github.com/IDEA-Research/GroundingDINO/releases/download/v0.1.0-alpha/groundingdino_swint_ogc.pth”
raw.githubusercontent.com · readme · Prepare pretrained checkpoints · retrieved 2026-10-15 -
Pre-training data of base model ram (not tracked here) (broad web data: may contain benchmark test sets) · split: unknown
Evidence
“Contextual Component Clustering (C 3 ) module id build upon visual foundation models such as RAM [37], Grounding DINO [24], and SAM [22]”
arxiv.org · paper · Section 4.3 · retrieved 2026-10-15
Benchmark cells
Industrial anomaly detection
Unknown MVTec AD · unknown for its base model OpenAI CLIP: training data not disclosed in enough detail
Unknown MVTec AD 2 · unknown for its base model OpenAI CLIP: training data not disclosed in enough detail
Unknown MVTec LOCO AD · unknown for its base model OpenAI CLIP: training data not disclosed in enough detail
Unknown VisA · unknown for its base model OpenAI CLIP: training data not disclosed in enough detail
Unknown Real-IAD · unknown for its base model OpenAI CLIP: training data not disclosed in enough detail
Unknown MPDD · unknown for its base model OpenAI CLIP: training data not disclosed in enough detail
Unknown BTAD · unknown for its base model OpenAI CLIP: training data not disclosed in enough detail
Unknown KolektorSDD2 · unknown for its base model OpenAI CLIP: training data not disclosed in enough detail
Notes
Training-free for all industrial/logical results (few-normal-shot: k normal reference images from the target training split at inference). The SAM used is HQ-SAM ViT-H (lkeab/hq-sam). A separate few-abnormal-shot variant with a trainable adapter is used only for medical datasets (Appendix C).
Sources
- UniVAD: A Training-free Unified Model for Few-shot Visual Anomaly Detection (CVPR 2025) · paper · retrieved 2026-10-15
- UniVAD README · readme · retrieved 2026-10-15
- UniVAD.py · code · retrieved 2026-10-15