kakaobrain / bassl Goto Github PK

Dear author:
Thanks for contributing an insightful work and make it published! it's helpful!
Your readme instruction file is really detailed. However, I can not find a "inference demo on single video" script . For a lot of researchers, A simple demo code will be the first step to try and play with your work. Thank you.

Hi,
Thanks for this wonderful work. Can you please release the setup and instruction for how to run on the BBC and the OVSD datasets?

Upon running scripts/download_movienet_data.sh, the annotation file is downloaded successfully but not the frames file. I get the following error:
unzip: cannot find or open ./bassl/data/movienet/movie1K.keyframes.240p.v1.zip, ./bassl/data/movienet/movie1K.keyframes.240p.v1.zip.zip or ./bassl/data/movienet/movie1K.keyframes.240p.v1.zip.ZIP. ls: cannot access './bassl/data/movienet/240P': No such file or directory

Thanks for your wonderful work!
I have a question about the jsons in data/movienet/anno folder, if I want to train on my own videos, how to prepare these files?
│─anno
│─ anno.pretrain.ndjson
│─ anno.trainvaltest.ndjson
│─ anno.train.ndjson
│─ anno.val.ndjson
│─ anno.test.ndjson
│─ vid2idx.json

Thanks a lot!!

Hello,
I read that you might be working on a demo on how to predict on a single video.
I was able to create my own dataloader and call trainer.predict() but the output is not binary (boundary or not boundary).
Does this model support scene boundary prediction (if so could you detail what are the steps? I just need to understand how i can make it work) or is it only a shot encoding model?

Thank you very much

Thanks for open sourcing your great work.

Found two hardcoded paths in bassl/finetune/utils/metric.py line181 and 186, causing not found error when this method is being called. Also, vid2ind.json and shot files are not included in anno.tar.

Great work!
Is there are any demo that you can provide to run inference on one's own videos?

Thanks for your wonderful work.

I would like to share a 3rd party support for BaSSL in our ModelScope, the largest Model Community in Chinese now, which also provide an online demo.
We hope that the link to demo of BaSSL in ModelScope can be added to the readme file.

Thanks again!

e.g.
What is the minimum GPU memory required for fine-tuning?
How long did it take to do fine-tuning?

I run the demo， but it need to install the shotdetect_scenedetect_lgss，I cannot find it

When I ran the pretrain code，loading anno.pretrain.ndjson raised this problem ‘json.decoder.JSONDecodeError: Expecting ',' delimiter: line 1 column 19268587 (char 19268586)’. Is there a format issue in the anno.pretrain.ndjson？

Many thanks for sharing your code.

I download anno.tar from https://arena.kakaocdn.net/brainrepo/bassl/data/anno.tar, and I can't find vid2idx.json, where can I get this file?

Thank you very much for your generous sharing, but I ran into some problems with the implementation.
During the data preparation stage, I tried to download the data with the command bash scripts/download_movienet_data.sh
but I failed, and got the following prompt：

How can I get the key-frames of MovieNet-SSeg dataset?

Thank you for your great work!

While I was following your setup instructions, the link provided in the script file for downloading keyframes is not working anymore. It seems that MovieNet is now downloadable by following the steps below after sign-up.

# Configure install
pip install opendatalab

odl login       # Login
odl get MovieNet        # Download full dataset

Hello, I found that during the pre-training process, the memory occupied keeps increasing in the iteration process, I want to know why this is, is the same for your training process and how much memory does it take to train an epoch? Thanks!