Closed yiminglin-ai closed 5 years ago
Hi Mark, I have been running the voc_convert.py for few days and there are 587703 images now. How many in total should I expect? Best, Yiming
Hi, I'm also kinda in trouble dealing with voc_convert.py for few days. I wonder if you could share the cropped dataset when done? Many thanks.
For an answer to your specific question about the total number of images in the dataset, there are approximately 10 million (there are that many annotations): https://research.google.com/youtube-bb/
voc_convert does have some issues with speed when attempting to convert the entire dataset at once. I would recommend editing the csv files to work with a subset of the dataset instead. For more details see here: https://github.com/mbuckler/youtube-bb/issues/14
If folks are interested in just the decoded images (not the video data) then they may be interested in a modified version of this repo which downloads, decodes, and then deletes the videos: https://github.com/mehdi-shiba/youtube-bb-utility
Hi Mark, I have been running the voc_convert.py for few days and there are 587703 images now. How many in total should I expect? Best, Yiming