gdwang08

@35454652 You may check the correct dataset you downloaded. It seems that the dataset you used has different numbers of frames or objects. You could check https://github.com/gdwang08/Jigsaw-VAD/issues/2 for reference.

@harveyzoka It seems that you do not have '0000.jpg', you may doule check your dataset directory. In case of incomplete Avenue dataset, I have uploaded a copy of the [dataset](https://drive.google.com/file/d/1LGAkgoqu5AQJzkqpR8s8R97xbXK5S9Mq/view?usp=share_link)...

@banzhuan37 The network is borrowed from the deep+wide version of the architecture in [SS-MTL](https://openaccess.thecvf.com/content/CVPR2021/papers/Georgescu_Anomaly_Detection_in_Video_via_Self-Supervised_and_Multi-Task_Learning_CVPR_2021_paper.pdf).