Awesome

IMPLICIT FOREGROUND-GUIDED NETWORK FOR ANOMALY DETECTION AND LOCALIZATION

This is the code for paper : IMPLICIT FOREGROUND-GUIDED NETWORK FOR ANOMALY DETECTION AND LOCALIZATION [ICASSP24']

Datasets

• VisA dataset
• BTAD dataset

Pretrained Models

We provide our model checkpoints to reproduce the performance report in the papar at : Baidu Drive (password:1mfd)

Evaluating

The test script requires :
 --gpu_id arguments
 --data_path the location of the VisA (or BTAD) anomaly detection dataset
 --checkpoint_path the folder where the checkpoint files are located

python test_IFgNet.py

Experimental Results

Visualization

Training

If you want to train a model from scratch, the train script requires :
 --gpu_id arguments
 --data_path the location of the VisA (or BTAD) anomaly detection dataset
 --anomaly_source_path the location of the DTD dataset

python train_IFgNet.py