POSE ESTIMATION API

---

This repository is API based Body and Hand Pose Estimation.

Really Simple :)

Just run the server and request to the server with any image file which contains body and hand(for hand body is required) That's all :)

Okay what is the workflow?

• You should install packages. (descripted inside of requirements.txt)
• Run the server for the requests
• Request to the server with image
• Get your response and Enjoy :)

How can you install packages?

Install PyTorch by following the quick start guide here (use pip)

For another requirements, just run command below.

$ cd pose_est_api/
$ pip3 install -r requirements.txt

For quick start you can install models from link below

dropbox

Download the pytorch models(.pth) and put them in a directory named model in the project root directory

After you have installed those packeges, what you will do next?

Now you can run your server first.

$ uvicorn post_est:app --reload

Then you should be get the messages below:

INFO:     Uvicorn running on http://127.0.0.1:8000 (Press CTRL+C to quit)
INFO:     Started reloader process [76894] using statreload
INFO:     Started server process [76896]
INFO:     Waiting for application startup.
INFO:     Application startup complete.

Let's try the server

For testing the server you can use req.jpg.

$ python3 main.py req.jpg

What is your response?

Do you want to save response image? [y/N]: y

After said yes, you get message below:

image saved as res.jpg

[IMPORTANT!]

• You can configure post json context from main.py.
• If you want to draw position keypoints of hand you should define value of "hand_pose" as 1.
• To draw rectangle for right and left hands(if it exists) you should define "which_hands" as 1.