Datasets:
Modalities:
Image
Languages:
English
Size:
100K<n<1M
Tags:
human action recognition
skeleton-based human action recognition
joint skeletons
human interaction
cyber-physical-social systems
digital twins
License:
John Martins
Reduced zipped file sizes to each below 500 MB, this way Huggingface can process them into parquet files using croissant and visualize them.
583c0cb Download data/train-020.zip from saluslab/HM-SYNC: direct link, hf CLI and curl.
- Browser
- Download file 485 MB
-
https://huggingface.co/datasets/saluslab/HM-SYNC/resolve/main/data/train-020.zip
- Command line
-
hf download hf://datasets/saluslab/HM-SYNC/data/train-020.zip
-
curl -L -o train-020.zip https://huggingface.co/datasets/saluslab/HM-SYNC/resolve/main/data/train-020.zip
485 MB
- Xet hash:
- fba9808a9c3366858388b4792f85af7c8867434c09ec8742ba3dc8b56d0637fe
- Size of remote file:
- 485 MB
- SHA256:
- 9b6e96b0a8bb9a17162dfce5addec1e23611c142aefae750d2a0cb8889ab92fd
·
Xet efficiently stores Large Files inside Git, intelligently splitting files into unique chunks and accelerating uploads and downloads. More info.