hangzhaomit / hacs-dataset Goto Github PK

For some of the videos I get the error:

Downloading video: https://www.youtube.com/watch?v=T11Rmq-r3h0
ERROR: Signature extraction failed: Traceback (most recent call last):
  File "/home/tobiascz/miniconda3/envs/kinetics/lib/python3.4/site-packages/youtube_dl/extractor/youtube.py", line 1342, in _decrypt_signature
    video_id, player_url, s
  File "/home/tobiascz/miniconda3/envs/kinetics/lib/python3.4/site-packages/youtube_dl/extractor/youtube.py", line 1250, in _extract_signature_function
    res = self._parse_sig_js(code)
  File "/home/tobiascz/miniconda3/envs/kinetics/lib/python3.4/site-packages/youtube_dl/extractor/youtube.py", line 1314, in _parse_sig_js
    jscode, 'Initial JS player signature function name', group='sig')
  File "/home/tobiascz/miniconda3/envs/kinetics/lib/python3.4/site-packages/youtube_dl/extractor/common.py", line 1004, in _search_regex
    raise RegexNotFoundError('Unable to extract %s' % _name)
youtube_dl.utils.RegexNotFoundError: Unable to extract Initial JS player signature function name; please report this issue on https://yt-dl.org/bug . Make sure you are using the latest version; see  https://yt-dl.org/update  on how to update. Be sure to call youtube-dl with the --verbose flag and include its complete output.
 (caused by RegexNotFoundError('Unable to extract Initial JS player signature function name; please report this issue on https://yt-dl.org/bug . Make sure you are using the latest version; see  https://yt-dl.org/update  on how to update. Be sure to call youtube-dl with the --verbose flag and include its complete output.',)); please report this issue on https://yt-dl.org/bug . Make sure you are using the latest version; see  https://yt-dl.org/update  on how to update. Be sure to call youtube-dl with the --verbose flag and include its complete output.
Error processing [Slacklining/T11Rmq-r3h0]: 'streams'

Iam using youtuble-dl 2019.05.20.
So far I am not sure where this is coming from.

Why is the actual length of some downloaded videos different from that in the annotation files

Hello, thank you for your great work. I was hoping to ask that I have submitted request for missing.txt but haven't received a response.

Hello, We are ready to use your HACS dataset to do fMRI research. We have completely followed your instructions in README and used official script to download videos in youtube. However, as some people have reflected before, our IP is blocked due to crawling too many videos, which seems like the youtube has upgraded anti-web crawling measures. We check the youtube-dl and it's the latest version.
You have mentioned that submitting a request with user agreement. Does it mean we need to finish the google form "HACS Video Request"? In that way, we need to supply almost the whole dataset as the missing video, and it is not allowed as the missing.txt is too large. We sincerely hope that there will be a reliable solution and use your dataset to do some research. Thanks for your help!

I would like to thank the authors for the great contribution of annotated videos/clips!

I see something weird, and not sure whether I understand correctly. I downloaded the videos and obtained the missing ones following instruction on the README.txt. Also I downloaded the json file which stores all the timestamp information of each video.

Somehow, I find the number of videos is much larger than what is annotated in json. For example, the folder "Drinking_beer" contains 1485 videos (mp4) in my download file, but the json file only has 202 titles that contain "Drinking_beer" in 'label'. This is for training set.

Do I understand the logic correctly? Or only a proportion of training videos (i.e., 202/1485) titles are annotated?

I noticed that the json file only contains 49581 rows. I guess maybe the video folder contains both video and clip, with the same folder name (for example "Drinking_beer") but different video_id names?

====Parsing clips====
[training set]: 492748 videos, 1509478 clips
[validation set]: 5981 videos, 20245 clips
[testing set]: 5987 videos, 20293 clips
====Parsing segments====
[training set]: 37613 videos
[validation set]: 5981 videos
[testing set]: 5987 videos
Done.

hi， thanks for your datasets.
I have submitted the form to request the downloading of the missing video, but there is currently no response.
Is it still available for downloading?

Hello, Mr.Zhao!
Here I have questions when I read your paper.
1.Could you tell me why you set negative clips and positive clips,why not just use the positive clips? Does the negative clips act as the ambiguous class?
2.The paper said that every video has 3 clips with the negative to positive ratio being roughly 1 to 2, but when I see the HACS_clips_v1.1.1 file, every video has only one clip. Could you tell me the reason, whether I misunderstand the description?
Thank you!

hi， thanks for your work. Is it still available for downloading?
I have submitted a request with the txt file of missing video and sent email, but there is no response.

Hi, thank you for hosting the great dataset. I have a quick question about the segment dataset.

I understand that there are multiple action segments in one video, but do all the segments belong to one class or multiple classes? I glanced through the .json file, it seems to me that all the segments in one video will have the same class label. Thank you.

Sometimes it appears httperror 429, it seem like the youtube has upgraded anti-web crawling measures and our IPs have been banned. We have use the latest version of youtube-dl. We can just download about 2000 videos at most, that means we can't use your website to download missing videos which are too much.

Hi, I submitted a missing video request on February 13, 2023. But after a week I don't get the link to the missing videos. Can you check if my request was submitted successfully? Thanks!

I'm in a region without direct access to Youtube,and unfortunately the remote server we're using doesn't have vpn. May I send a request for the direct link for the whole video dataset ,which is split into multiple missing text ,following the submit instruction you have given?

Are there any video dataset link?
Problems will appear if we want to get access to YOUTUBE at certain region in the world :(

Hello there. I completely follow your instructions in README. But I run into a problem. I want to know if your method is to crawl video from youtube one by one according to the video ID.If so, what can i do when my IP is blocked due to crawling too many videos?

Thank you for your attention.
I have downloaded all of segments videos. When I was extracting frames from them, six videos among them seemed to be damaged, while there is no problem with the other videos. So I sumbitted a missing.txt for re-downloading the six videos:

the video ids are v_2wE42p2P85g, v_EKhVsJ3rtyM, v_PXS1LtrnQXI, v_qh0kaLdrzJU, v_VubOnSthyKQ, v_Wj40pGUd368.
the newly downloaded ones encounter the same problem. When I try to extract frames from these videos, I see this:

And if I try to watch this video, I see this:

the popup means that "This file cannot be played. This may be due to unsupported file types, incorrect file extensions, or corrupted files."
Could you have any advice for me? thank you very much.