AnomalyGPT: Detecting Industrial Anomalies using Large Vision-Language Models
๐ Project Page โข ๐ค Online Demo โข ๐ Paper โข ๐ค Model โข ๐น Video
Zhaopeng Gu, Bingke Zhu, Guibo Zhu, Yingying Chen, Ming Tang, Jinqiao Wang
Catalogue:
1. Introduction: [Back to Top]
AnomalyGPT is the first Large Vision-Language Model (LVLM) based Industrial Anomaly Detection (IAD) method that can detect anomalies in industrial images without the need for manually specified thresholds. Existing IAD methods can only provide anomaly scores and need manually threshold setting, while existing LVLMs cannot detect anomalies in the image. AnomalyGPT can not only indicate the presence and location of anomaly but also provide information about the image.
We leverage a pre-trained image encoder and a Large Language Model (LLM) to align IAD images and their corresponding textual descriptions via simulated anomaly data. We employ a lightweight, visual-textual feature-matching-based image decoder to obtain localization result, and design a prompt learner to provide fine-grained semantic to LLM and fine-tune the LVLM using prompt embeddings. Our method can also detect anomalies for previously unseen items with few normal sample provided.
2. Running AnomalyGPT Demo [Back to Top]
2.1 Environment Installation
Clone the repository locally:
git clone https://github.com/CASIA-IVA-Lab/AnomalyGPT.git
Install the required packages:
pip install -r requirements.txt
2.2 Prepare ImageBind Checkpoint:
You can download the pre-trained ImageBind model using this link. After downloading, put the downloaded file (imagebind_huge.pth) in [./pretrained_ckpt/imagebind_ckpt/] directory.
2.3 Prepare Vicuna Checkpoint:
To prepare the pre-trained Vicuna model, please follow the instructions provided [here].
2.4 Prepare Delta Weights of AnomalyGPT:
We use the pre-trained parameters from PandaGPT to initialize our model. You can get the weights of PandaGPT trained with different strategies in the table below. In our experiments and online demo, we use the Vicuna-7B and openllmplayground/pandagpt_7b_max_len_1024 due to the limitation of computation resource. Better results are expected if switching to Vicuna-13B.
| Base Language Model | Maximum Sequence Length | Huggingface Delta Weights Address |
|---|---|---|
| Vicuna-7B (version 0) | 512 | openllmplayground/pandagpt_7b_max_len_512 |
| Vicuna-7B (version 0) | 1024 | openllmplayground/pandagpt_7b_max_len_1024 |
| Vicuna-13B (version 0) | 256 | openllmplayground/pandagpt_13b_max_len_256 |
| Vicuna-13B (version 0) | 400 | openllmplayground/pandagpt_13b_max_len_400 |
Please put the downloaded 7B/13B delta weights file (pytorch_model.pt) in the ./pretrained_ckpt/pandagpt_ckpt/7b/ or ./pretrained_ckpt/pandagpt_ckpt/13b/ directory.
After that, you can download AnomalyGPT weights from the table below.
| Setup and Datasets | Weights Address |
|---|---|
| Unsupervised on MVTec-AD | AnomalyGPT/train_mvtec |
| Unsupervised on VisA | AnomalyGPT/train_visa |
| Supervised on MVTec-AD, VisA, MVTec-LOCO-AD and CrackForest | AnomalyGPT/train_supervised |
After downloading, put the AnomalyGPT weights in the ./code/ckpt/ directory.
In our online demo, we use the supervised setting as our default model to attain an enhanced user experience. You can also try other weights locally.
2.5. Deploying Demo
Upon completion of previous steps, you can run the demo locally as
cd ./code/
python web_demo.py3. Train Your Own AnomalyGPT [Back to Top]
Prerequisites: Before training the model, making sure the environment is properly installed and the checkpoints of ImageBind, Vicuna and PandaGPT are downloaded.
3.1 Data Preparation:
You can download MVTec-AD dataset from [this link] and VisA from [this link]. You can also download pre-training data of PandaGPT from [here]. After downloading, put the data in the [./data] directory.
The directory of [./data] should look like:
data
|---pandagpt4_visual_instruction_data.json
|---images
|-----|-- ...
|---mvtec_anomaly_detection
|-----|-- bottle
|-----|-----|----- ground_truth
|-----|-----|----- test
|-----|-----|----- train
|-----|-- capsule
|-----|-- ...
|----VisA
|-----|-- split_csv
|-----|-----|--- 1cls.csv
|-----|-----|--- ...
|-----|-- candle
|-----|-----|--- Data
|-----|-----|-----|----- Images
|-----|-----|-----|--------|------ Anomaly
|-----|-----|-----|--------|------ Normal
|-----|-----|-----|----- Masks
|-----|-----|-----|--------|------ Anomaly
|-----|-----|--- image_anno.csv
|-----|-- capsules
|-----|-----|----- ...
3.2 Training Configurations
The table below show the training hyperparameters used in our experiments. The hyperparameters are selected based on the constrain of our computational resources, i.e. 2 x RTX3090 GPUs.
| Base Language Model | Epoch Number | Batch Size | Learning Rate | Maximum Length |
|---|---|---|---|---|
| Vicuna-7B | 50 | 16 | 1e-3 | 1024 |
3.3 Training AnomalyGPT
To train AnomalyGPT on MVTec-AD dataset, please run the following commands:
cd ./code
bash ./scripts/train_mvtec.shThe key arguments of the training script are as follows:
--data_path: The data path for the json filepandagpt4_visual_instruction_data.json.--image_root_path: The root path for training images of PandaGPT.--imagebind_ckpt_path: The path of ImageBind checkpoint.--vicuna_ckpt_path: The directory that saves the pre-trained Vicuna checkpoints.--max_tgt_len: The maximum sequence length of training instances.--save_path: The directory which saves the trained delta weights. This directory will be automatically created.--log_path: The directory which saves the log. This directory will be automatically created.
Note that the epoch number can be set in the epochs argument at ./code/config/openllama_peft.yaml file and the learning rate can be set in ./code/dsconfig/openllama_peft_stage_1.json
4. Examples
An image of concrete with crack.
A crack capsule.
An image of a cut hazelnut.
A damaged bottle.
A photo of normal carpet.
A photo of a piece of wood with defect.
A piece of normal fabric.
License
AnomalyGPT is licensed under the CC BY-NC-SA 4.0 license.
Citation:
If you found AnomalyGPT useful in your research or applications, please kindly cite using the following BibTeX:
@article{gu2023anomalyagpt,
title={AnomalyGPT: Detecting Industrial Anomalies using Large Vision-Language Models},
author={Gu, Zhaopeng and Zhu, Bingke and Zhu, Guibo and Chen, Yingying and Tang, Ming and Wang, Jinqiao},
journal={arXiv preprint arXiv:2308.15366},
year={2023}
}
Acknowledgments:
We borrow some codes and the pre-trained weights from PandaGPT. Thanks for their wonderful work!