search

Prasad9 / Classify-HandGesturePose

Network to classify hand gesture poses.
GNU General Public License v2.0
129 stars 30 forks source link

Example of pose.csv for SVM solution #5

Open felagund89 opened 5 years ago

felagund89 commented 5 years ago

Hi Prasad, thanks for this repository!! Can you give us an example of pose.csv file for the SVM solution ? I dont understand how the csv is composed. Thanks. Best regards.

Antonio

Prasad9 commented 5 years ago

Hi Antonio,

Thanks for checking out the repository. The CSV file comprises of rows of 21 keypoints (or landmarks) coordinates of hand. Each of the keypoint is a 3D coordinate (x, y, z). So each of the row (or reading) will have 63 numbers in it.

Collect each type of pose in different CSV files. You are free to arrange the keypoint data in any order, but ensure that you maintain same order in all the files. Then, in the command line instruction, you pass these csv files as python pose/training/SVM.py './pose1.csv, ./pose2.csv, ./pose3.csv'

If you would like to generate these files through a tool written by me, please look into instructions mentioned in Neural network part.

felagund89 commented 5 years ago

Hi Antonio,

Thanks for checking out the repository. The CSV file comprises of rows of 21 keypoints (or landmarks) coordinates of hand. Each of the keypoint is a 3D coordinate (x, y, z). So each of the row (or reading) will have 63 numbers in it.

Collect each type of pose in different CSV files. You are free to arrange the keypoint data in any order, but ensure that you maintain same order in all the files. Then, in the command line instruction, you pass these csv files as python pose/training/SVM.py './pose1.csv, ./pose2.csv, ./pose3.csv'

If you would like to generate these files through a tool written by me, please look into instructions mentioned in Neural network part.

Ok thanks, so each row in a csv represent a video and each column represent a determinate keypoint in all the frame of the video?Thanks for your time.

Antonio

Prasad9 commented 5 years ago

Just a small clarification, each row corresponds to reading obtained from a particular frame in the video. If you are using the tool to generate a new csv file, and your input video file has 1000 frames in it out of which only 600 frames are able to form the pose, the csv file will be having 600 rows in it.