cannot find HEVI_data/fol_data

[linchunmian]

Hi, author. I have downloaded the HEVI dataset, and prepare to train the model. But I am confused about the dataset definition in the config.yaml. For instance, in 'fol_ego_train.yaml', what the data_root mean? When I check my HEVI dataset, I don't find the 'fol_data'. Please help me! Thanks.

[MoonBlvd]

Hi @ChunmianLin please follow the README file to download the extracted HEVI features for training and testing. The data_root is just where you save those features.

[linchunmian]

@MoonBlvd Thanks. But I still cannot get your mean fully. My HEVI folder is as follows, and the 'fol_ego_train.yaml' are as follow. My problem is how to match the folder name to the 'data_root' and 'ego_data_root' path Thanks in advance!

[MoonBlvd]

@ChunmianLin Please download the featrues through the two links in my readme file, these are 'train' and 'val' folders containing many '.pkl' files. Then the 'data_root' should point to where you save these featrues.

[MoonBlvd]

@ChunmianLin I remember the 'ego_data_root' is not used for training because those .pkl files already have the exrtacted ego_motion in it.

[linchunmian]

@MoonBlvd I got it, and the training is work on .pkl feature files you provided. So many thanks for your careful instructions. BTW, what and where is the 'ego_data_root' used for? If the training process of ego model finished, should I follow the step to run on test set in readme.file? I am newer in this area, and so sorry for disturbing your time frequently.

[MoonBlvd]

@ChunmianLin No worries. If I remember it correctly, for training/testing FOL on HEV-I dataset, you can use the pkl files we provided. But for testing on A3D dataset for anomaly detection, you will need to assigne your ego_data directory to the ego_data_root parameter.

[linchunmian]

@MoonBlvd Thanks for your help, again. I would prepare to continue the following step. Simultaneously, I am downloading A3D data from youtube, but partial videos cannot be pulled down due to the bug message 'youtube_dl.utils.DownloadError(ssl protocol)'. I see the completed A3D.pkl file you provided, but only several anomaly detection data in my local. Does it effect the testing?

[MoonBlvd]

You can always test on part of the dataset, shouldn't be problematic. The error you saw is probably a cookie problem, my other repo for the DoTA dataset has a solution that might help solving this issue: https://github.com/MoonBlvd/Detection-of-Traffic-Anomaly

[linchunmian]

Thanks. I encounter new problem, and could you please explain the reason why uses these features?

[MoonBlvd]

Not sure what your question is. These are the bounding box, flow and ego motion.

[linchunmian]

@MoonBlvd Sorry, I forget to unload the screenshot. As follow, I mean what is the function of these features.

[khoatcao]

@MoonBlvd. Hi I would love to receive your help on my issues. I have following instructions above and download links from the readme file. Link 1 : https://drive.google.com/file/d/1TE-smXm4dD2QgoCQHYmzLHqSsltoIxbe/view Link 2 : https://drive.google.com/file/d/1Vcu6NU7PwDOPTv6RU_7AuBfj6I0rj4dR/view

These two files were saved in my local as the following images:

https://user-images.githubusercontent.com/50226654/141648849-5043ee3d-ebd2-4c39-8898-3d3fb82dc750.png)

The errors images.

https://user-images.githubusercontent.com/50226654/141648537-32419b60-0451-4602-9acd-7d93a04f395c.png).

When I run the command :

python train_fol.py --load_config config/fol_ego_train.yaml.

The error had been raising in the image attached above. I hope that you can take a little time to review my jobs and helps me figure its out . Thanks.