The YCB-Video dataset contributed by PoseCNN is based on the YCB dataset. 21 objects are selected, and photographed with an RGBD camera to make 92 videos. The entire data set contains 133827 frames.
The YCB-Video dataset is a little hard to download from Google Drive. Because it only consists of a super large compressed file (256GB) and does not support resumable transmission. Once the download fails, you will face an rate limitation (usually 24 hours) and have to download again from the first byte!😭
So I used sub-volume compression, each file is round 10GB in size with a verifiable hash. Although multi-threaded downloading is still not allowed, at least you can re-download from a sub-volume instead of the whole. Furthermore, I added text labels for YOLO training to each image under the data folder.
YOLO Demos
The weights file can be downloaded from here https://git.chengke.name/chengke-codes/yolo_ycb
Download Links
Thanks a lot for sharing the dataset in a downloadable way.
One more point
I think one information that is missing from the page is how to merge the splits. In order to merge the split after completion of download, run this command:
7z x -tsplit YCB_dataset.7z.001
Thank you for loving this.
I don’t understand the command.
The files in sharepoint/onedrive are not ‘YCB_dataset.7z.001’ but ‘ycb_dataset.z01’
Could you please provide me the code you used to unzip?
I appreciate your help
Joan
I uploaded zip volumes, so the 7z decompression command is no longer needed…
Hi Chengke,
The download link of the sub-volume compression no longer exists. Can you please send me another link please? I will be grateful if you can share it with me.
Thank you so much in advance,
Regards.
I just checked download links of onedrive are working. Could you try it again?
Yes Now it is working 🙂 . Thank you for your feedback ^^
Dear Dr. Chengke,
I hope you do not mind me getting in touch. I am a master’s student working on a draft for object recognition. The video results you show on youtube are promising. I wonder if you can share the inference Pytorch code of Yolov5 or the weights of your model. I would be grateful if you could share them with us.
I am looking forward to your positive reply.
Yours sincerely.
I just uploaded them to here https://github.com/chengke-codes-sk/ycb_demo/tree/master
Hello,
Can you provide the weights of your models (yolov4 and yolov5) trained on the YCB-Video dataset? Please. Thank you.
Please see the last comment 🙂
Hello,
Are the Yolov5 weights also trained to detect object masks? Thank you.
Also, the weights and the videos from your github repo can’t be downloaded anymore due to the bandwidth limit being reached
I migrated the repository to https://git.chengke.name/chengke-codes/yolo_ycb without the bandwidth limitation.
I think they are only trained to detect bounding box.
Hi! The onedrive link is not working
Hello, I think the link was suspended due to over bandwidth limitation of Onedrive. Now it is working.
Dear Dr. Chengke,
I am a master’s student working about 6D object recognition. Thanks for your great work. I wonder if you can share the training Pytorch code of Yolov5 of your model. I would be grateful if you could share it with us.
Looking forward to your positive reply.
Yours sincerely.
Sorry I don’t think I have them… However, you could try to check config files of the yolov5 directory. there should be some interesting parameters to help you.
Ok. Thank you for sharing again. The last question I want to ask is which version of Yolov5 did you use?
I don’t remember the accurate version number. However I trained these weight files 2 years ago. Thus you could find a reasonable version for that time…
Hi,
I have downloaded the YCB from your ondrive, but while I unzip ./ycb_dataset.zip, it meeat “unzip bad zipfile offset (local header sig)”. The md5 are the same as yours. Could you please tell me how you unzip the files please?
Hello,
Did you download all part files?
Thank you for your reply! I fix it buy this comand “zip -F ./ycb_dataset.zip –out ./ycb_dataset_new.zip”
that is great!