ekosman/AnomalyDetectionCVPR2018-Pytorch

Complete Execution steps involved in this project

Chuttyboy opened this issue · 1 comments

Chuttyboy commented

AnomalyDetectionCVPR2018-Pytorch

Pytorch version of - https://github.com/WaqasSultani/AnomalyDetectionCVPR2018

Known Issues:

  • AUC is not exactly as reported in the paper (0.70 vs 0.75) - might be affected by the weights of C3D

  • Error happens while using python3 with anaconda ( I recommend you to use pipenv )

C3D Weights

I couldn't upload here the weights for the C3D model because the file is too big, but it can be found here:
https://github.com/DavideA/c3d-pytorch

Dataset (Optional)

Can be downloaded from:
https://www.dropbox.com/sh/75v5ehq4cdg5g5g/AABvnJSwZI7zXb8_myBA0CLHa?dl=0

Download it and save it as a " data " folder

Precomputed Features

Can be downloaded from:
https://drive.google.com/drive/folders/1rhOuAdUqyJU4hXIhToUnh5XVvYjQiN50?usp=sharing

Download and save it as " anomaly_features "

Execution steps :

1.Feature Extraction :

way 1: Download dataset from the above-given link

way 2: Extracting feature from dataset to run this command
Arguments:

  • dataset_path - path to the directory containing videos to extract features for (the dataset is available for download above)
  • model_type - which type of model to use for feature extraction (necessary in order to choose the correct pre-processing)
  • pretrained_3d - path to the 3D model to use for feature extraction

python feature_extractor.py --dataset_path "path-to-dataset" --model_type "fe-model-eg-c3d" --pretrained_3d "path-to-pretrained-fe"

Run this command :

python feature_extractor.py --dataset_path data --model_type c3d --pretrained_3d pretrained/c3d.pickle

2.Training :

Arguments:

  • features_path - path to the directory containing the extracted features (pre-computed features are available for download above, or supply your own features extracted from the previous stage)
  • annotation_path - path to the annotations file (Available in this repository as Train_annotations.txt

python TrainingAnomalyDetector_public.py --features_path "path-to-dataset" --annotation_path "path-to-train-annos"

Run this command :

python TrainingAnomalyDetector_public.py --features_path anomaly_features --annotation_path Train_Annotation.txt

ekosman commented

Closing this as it is a copy of README.md