PROPOSE

PRObabilistic POSE estimation

Getting started

Install the package from source:

pip install git+https://github.com/sinzlab/propose.git

Loading a pretrained cGNF

We provide the pretrained model which you can load with the following code snippet.

from propose.models.flows import CondGraphFlow

flow = CondGraphFlow.from_pretrained('ppierzc/cgnf/cgnf_human36m:best')

HRNet Loading

You can also load a pretrained HRNet model.

from propose.models.detectors import HRNet

hrnet = HRNet.from_pretrained('ppierzc/cgnf/hrnet:v0')

This will load the HRNet model provided in the repo. The model loaded here is the pose_hrnet_w32_256x256 trained on the MPII dataset.

Requirements

Requirements for the package

The requirements for the package can be found in the requirements.txt.

Docker

Alternatively, you can use Docker to run the package. This project requires that you have the following installed:

• docker
• docker-compose

Ensure that you have the base image pulled from the Docker Hub. You can get the base image by running the following command:

docker pull sinzlab/pytorch:v3.9-torch1.9.0-cuda11.1-dj0.12.7

Running notebooks

1. Clone the repository.
2. Navigate to the project directory.
3. Rundocker-compose build base
4. Rundocker-compose run -d -p 10101:8888 notebook_server
5. You can now open JupyterLab in your browser at http://localhost:10101.

Available Models

Model Name	description	Artifact path	Import Code
cGNF Human 3.6m	Model trained on the Human 3.6M dataset with MPII input keypoints.	ppierzc/propose_human36m/mpii-prod:best	from propose.models.flows import CondGraphFlow
HRNet	Instance of the official HRNet model trained on the MPII dataset with w32 and 256x256	ppierzc/cgnf/hrnet:v0	from propose.models.detectors import HRNet

Run Tests

To run the tests, from the root directory call:

docker-compose run pytest tests

Note: This will create a separate image from the base service.

Data

Rat7m

You can download the Rat 7M dataset from here.

Human3.6M dataset

Due to license restrictions, the dataset is not included in the repository. You can download it from the official website.