Skip to content
PyTorch implementation of the Dec 2018 paper
Jupyter Notebook Python
Branch: master
Clone or download
Latest commit 34ef01d May 24, 2019
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
.ipynb_checkpoints added updated code with model weights 368x368 and 184x184 May 20, 2019
model-results code commit May 1, 2019
test-images/office code commit May 1, 2019
.gitignore adding images for readme May 1, 2019
10.png added updated code with model weights 368x368 and 184x184 May 20, 2019
12.png added updated code with model weights 368x368 and 184x184 May 20, 2019
15.png added updated code with model weights 368x368 and 184x184 May 20, 2019
CONFIG.py added updated code with model weights 368x368 and 184x184 May 20, 2019
CONSTANTS.py added updated code with model weights 368x368 and 184x184 May 20, 2019
LICENSE Initial commit May 1, 2019
README.md Update README.md May 23, 2019
custom_transforms.py added updated code with model weights 368x368 and 184x184 May 20, 2019
dataset.py added updated code with model weights 368x368 and 184x184 May 20, 2019
estimate-pose.ipynb
gaussian_15X15_sigma_7.npy code commit May 1, 2019
gaussian_5X5_sigma_3.npy added updated code with model weights 368x368 and 184x184 May 20, 2019
gaussian_9X9_sigma_3.npy code commit May 1, 2019
im_ids.npy code commit May 1, 2019
img_id_to_annotations.npy code commit May 1, 2019
img_id_to_image_info.npy code commit May 1, 2019
matching-res.png code commit May 1, 2019
model-wts-184.ckpt added updated code with model weights 368x368 and 184x184 May 20, 2019
model-wts-368.ckpt added updated code with model weights 368x368 and 184x184 May 20, 2019
model_utils.py added updated code with model weights 368x368 and 184x184 May 20, 2019
net-arch.png adding images for readme May 1, 2019
nn_model.py added updated code with model weights 368x368 and 184x184 May 20, 2019
plot_utils.py
preds.png code commit May 1, 2019
sample-trn-img.png
train2014 code commit May 1, 2019
trn-hms.png code commit May 1, 2019
trn-img-heatmaps.png code commit May 1, 2019
trn-img-paf-vectors.png code commit May 1, 2019
trn-pafs.png code commit May 1, 2019

README.md

PyTorch-OpenPose-Realtime-Multi-Person-2D-Pose-Estimation-using-Part-Affinity-Fields

output.gif: https://oregonstate.box.com/s/2s390ub7b5880o4qnrq3io5xdmxhjsxe

PyTorch implementation of the Dec 2018 paper: https://arxiv.org/abs/1812.08008

Go through estimate-pose.ipynb for training and evaluation code on sample image, video.

Model has been trained on MS-COCO 2014 dataset on 368x368 and 184x184 resolutions (model-wts-368.ckpt, model-wts-184.ckpt).

Has additional PAF's from Shoulder->Wrist and Hip->Ankle for improved matching in crowded scenes.

You can change the threholds for PAF map o/p values, Heatmap threshold and part matching threshold in CONFIG.py (for more, less conf joint preds vs less, more confident).

Part Matching formulation uses Munkres for one-one least cost matching.


Model Results:

Joint Matching: (With the help of predicted Part Affinity Field vectors)

Network Architecture: (1st 10 layers from VGG-16 as backbone(F), 4 PAF stages(L), 2 Heatmap stages(S))

Training Image, Generated PAF's, Generated Joint Heatmaps:

You can’t perform that action at this time.