ninatu/anomaly_detection

This is the official implementation of "Anomaly Detection with Deep Perceptual Autoencoders".

anomaly_detection's Issues

CAMELYON16 dataset

I am having a hard time downloading the CAMELYON16 dataset. Can anyone please provide a source from where you could download the dataset? Thanks.

How to use dpa with the original dataset

Hello.
Thank you very much for sharing your admirable study to the public!
I would like to train and test my pathological images to detect anomalies by using your dpa model.
Would you mind letting me know where should I put my train, validation, test data and several anomaly images to choose hyperparameter, and how config file should be changed, to use your dpa?
Thank you,

ModuleNotFoundError: No module named 'pretrainedmodels'

I encountered this error while running deep_if model. Can you please add this module?

training and evaluating Camelyon16 dataset

Hi,

Thanks you once again for your wonderful work and help that you did last time (almost a year ago) in downloading the Camelyon16 dataset. After a long break, I am once again trying to play a bit with your work. I downloaded and have preprocessed the Camelyon16 dataset. Now my goal is to run your perceptual autoencoder to find the anomalies.

However, I am not sure how I can do that and therefore, I need your guidance on this.
The main issue is I guess I need to know which config file I should use. I couldn't find any config for Camelyon16 in "anomaly_detection/configs/dpa" folder. However, there are several in "anomaly_detection/configs/dpa/camelyon16/final". Could you tell me what "class_healthy" folder and all the subfolders, i.e., run_0, run_1, run_2 actually means?

What I want to do is to run a bunch of test images and get the result that shows which images are anomalies among these.
Your help on this would be highly appreciated.
Thanks in advance.

About camelyon16 preprocessing

Hi,
Here may be a bug ,
In Line 28 of camelyon16_preprocessing/scripts/1_convert_annotation_to_json.py,
X = list(map(lambda x: float(x.get('x')) should be changed to
X = list(map(lambda x: float(x.get('X'))
Best wishes

Validation set for the early stopping

Hi,

Thanks for your great work! I have a question about the early stopping. According to your paper, 200 tumor patches (abnormal samples) were used as the validation set. In your code, I noticed that the sum of the adversarial loss and reconstruction loss of the validation set was used as the sign for early stopping. I am a little confused why the loss score of abnormal samples in the validation set instead of the AUC score of validation sets with both abnormal and normal images available is used for early stopping.

I am looking forward to your reply. Thank you!