Contamination Map Methodology

Vision / Model

AdaCLIP, trained on all 14 datasets (demo model)

Organisation
Huazhong University of Science and Technology et al.
Family
AdaCLIP · version pretrained_all.pth
Released
2024-07-22 (arXiv v1 submission date (https://arxiv.org/abs/2407.15795); release date of this checkpoint not stated)
Checkpoints
weights/pretrained_all.pth
Weights
Open · MIT
Links
PaperCode

Training data disclosure: Full list

Level determined for 50% of benchmark cells.

“To construct a robust ZSAD model for demonstration, we also train our AdaCLIP on all AD datasets mentioned above:”
raw.githubusercontent.com · readme · Train · retrieved 2026-10-15
“br35h brain_mri btad clinicdb colondb”
raw.githubusercontent.com · readme · Train (command line, first line of --training_data) · retrieved 2026-10-15
“dagm dtd headct isic mpdd mvtec sdd tn3k visa”
raw.githubusercontent.com · readme · Train (command line, second line of --training_data) · retrieved 2026-10-15

Built on

Everything these base models may have seen is inherited by AdaCLIP, trained on all 14 datasets (demo model).

Other declared training data

Benchmark cells

Industrial anomaly detection

Test data seen MVTec AD · in fine-tuning data: its test split was used for training

Evidence (3)
“br35h brain_mri btad clinicdb colondb”
raw.githubusercontent.com · readme · Train (command line, first line of --training_data) · retrieved 2026-10-15
“dagm dtd headct isic mpdd mvtec sdd tn3k visa”
raw.githubusercontent.com · readme · Train (command line, second line of --training_data) · retrieved 2026-10-15
“Only utilize the test dataset for both training and testing”
raw.githubusercontent.com · code · dataset/base_dataset.py · retrieved 2026-10-15

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

Test data seen VisA · in fine-tuning data: its test split was used for training

Evidence (3)
“br35h brain_mri btad clinicdb colondb”
raw.githubusercontent.com · readme · Train (command line, first line of --training_data) · retrieved 2026-10-15
“dagm dtd headct isic mpdd mvtec sdd tn3k visa”
raw.githubusercontent.com · readme · Train (command line, second line of --training_data) · retrieved 2026-10-15
“Only utilize the test dataset for both training and testing”
raw.githubusercontent.com · code · dataset/base_dataset.py · retrieved 2026-10-15

Unknown Real-IAD · unknown for its base model OpenAI CLIP: training data not disclosed in enough detail

Test data seen MPDD · in fine-tuning data: its test split was used for training

Evidence (3)
“br35h brain_mri btad clinicdb colondb”
raw.githubusercontent.com · readme · Train (command line, first line of --training_data) · retrieved 2026-10-15
“dagm dtd headct isic mpdd mvtec sdd tn3k visa”
raw.githubusercontent.com · readme · Train (command line, second line of --training_data) · retrieved 2026-10-15
“Only utilize the test dataset for both training and testing”
raw.githubusercontent.com · code · dataset/base_dataset.py · retrieved 2026-10-15

Test data seen BTAD · in fine-tuning data: its test split was used for training

Evidence (3)
“br35h brain_mri btad clinicdb colondb”
raw.githubusercontent.com · readme · Train (command line, first line of --training_data) · retrieved 2026-10-15
“dagm dtd headct isic mpdd mvtec sdd tn3k visa”
raw.githubusercontent.com · readme · Train (command line, second line of --training_data) · retrieved 2026-10-15
“Only utilize the test dataset for both training and testing”
raw.githubusercontent.com · code · dataset/base_dataset.py · retrieved 2026-10-15

Unknown KolektorSDD2 · unknown for its base model OpenAI CLIP: training data not disclosed in enough detail

Notes

Demo checkpoint (HF Space default option 'ALL'), not used for paper results. Trained on the test splits (per dataset code) of all 14 industrial and medical datasets, so it has seen the test sets of MVTec AD, VisA, MPDD and BTAD; any zero-shot evaluation of it on these is contaminated. 'sdd' is KolektorSDD (KSDD v1), not KSDD2.

Sources