file missing for training stlt

[patrolli]

hello, I tried to train the stlt model but got some errors. I first run create_dataset.py to prepare dataset, after that, I canot find the videoid2size.json file, which however is necessary for running the code. How can I get this json file? Thanks a lot ~

[gorjanradevski]

Good catch. You can download the .json file from this link: https://drive.google.com/file/d/1ANaDAxXoA63CA9zXalnmaskqfO4cftW4/view?usp=sharing

Let me know if you encounter any other issues!

[patrolli]

Thanks! I encountered another problem but not sure is caused by this repo. When I finish running the create_dataset.py code, the dump size of the result json file is 0, as shown in figure.

I am not sure what is caused by, could you please provide the resutl json file? Thanks a lot ~

[gorjanradevski]

I just tested everything and it should work. What I did:

• Downloaded train.json and validation.json in data/ from here.
• mkdir data/comp_annotations.
• Downloaded detected_compositional_part1/2/3.json in data/comp_annotations from here .
• Run create_datasets.py:

python src/create_datasets.py --train_data_path "data/train.json" --val_data_path "data/validation.json" --annotations_path "data/comp_annotations/"

the train_dataset.json and val_dataset.json should be dumped inside the data/ directory.

ps. I updated the README.md with better data preparation instructions and pushed small fix in create_datasets.py (open() missing inside json.load).

Let me know how it goes!

[gorjanradevski]

I'm closing this issue, feel free to reopen if the problem persists.