convert video labels from vatic.js to input for openpose or deeplabcut
- Go from video data to neural network inputs (images and keypoint labels) for pose estimation packages for transfer learning with openpose or Deeplabcut
- Upload video to vatic.js. Code can be run locally or here. Video example files in ./files/video.
- Label video, including bounding box around the figure. Can upload example frame zip file in ./files/frame_zip and xml file in ./files/xml to see labels
- Download output files. Vatic produces a zip file containing frames from the video and an xml file containing the labels. Unzip frame zip file. Examples in ./files/xml and ./files/image
- To get openpose .json file and resized images, run notebook
xml_to_openposeNNinput.ipynb. Change notebook for use case. - To get deeplabcut pandas dataframe and resized images, run notebook
xml_to_DLCNNinput.ipynb. Change notebook for use case.