Skip to content
Model to classify yoga pose type and estimate joint positions of a person from an image
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
COCO trained model as per additional pafs, inference code in estimate-pose… May 20, 2019
classification-model
dataset-creation finished training classifier model, overfits ds gets 97% accuracy Feb 11, 2019
papers added ref paper for lightweight openpose model Feb 2, 2019
pose-estimation-model changes to COCO training model notebook, added results from pre exist… Mar 21, 2019
user-stories
week-1 added week1 to read list Jan 20, 2019
.gitignore
README.md Update README.md Mar 21, 2019
half-lotus-res-1.png tested a pre existing model, added sample results Jan 24, 2019
half-lotus1.jpg
half-lotus2.jpg tested a pre existing model, added sample results Jan 24, 2019
matching-results.png Added final joint matching code, added sample results Apr 29, 2019
picture_demo.py tested a pre existing model, added sample results Jan 24, 2019
predict-joints(pose-estimation).jpg adding initial files and user stories Jan 20, 2019
references.txt tested a pre existing model, added sample results Jan 24, 2019
result-2.png tested a pre existing model, added sample results Jan 24, 2019
result.png

README.md

yoga-pose-estimation

Model to classify yoga pose type and estimate joint positions of a person from an image

Pose Estimation part: alt text

  • Get join estimation model, training params from OpenPose (Done)

  • OpenPose paper : https://arxiv.org/abs/1812.08008

  • Earlier paper : https://arxiv.org/abs/1611.08050

  • Download Microsoft's COCO dataset for labeled keypoints. (17 Joints and 38 limb connections) (Done)

  • COCO : http://cocodataset.org/#download

  • Write code to calculate ground truth joint Heatmaps and PAF's (Part affinity fields) from labeled keypoints (Done)

  • Write Dataset class, custom transforms and vizualizations plots (Done)

  • Write the NN Model to be trained in stages, implement from original paper (Done)

  • Write Loss calculation part (Done)

  • Train model (PAF Stages 1-4) (Done)

  • Train model (Heatmaps Stages 1-2) (Done)

In Progress:

  • Write the inference logic, nms to connect joints from prdicted pafs, heatmaps.

Sample Training Image: alt text

PAFs: alt text

Heatmaps: alt text

You can’t perform that action at this time.