Which video format do it take and do it give output in the form of json?

[arshikasoni]

How can we take output in U-V format? and why is it giving output of just last frame

[trrahul]

Could you post the output log? You can change the ffmpeg params to get the output you want. https://github.com/trrahul/densepose-video/blob/2157512c7604da1f00ef9a2065d4c81deff382df/infer_vid.py#L145

[arshikasoni]

I am running in colab so here is the output

also I am getting just the last frame of the video like as shown in the above video, I want to know how can we get some u-v cordinates of some body parts from where we can calculate angles of some body parts.

[trrahul]

You can see it on the terminal.

[arshikasoni]

Can't we do something like we get some kind of json or csv of the body coordinates, so that we can calculate angles between body parts?

[trrahul]

You can see the output logs in stdout, normally in terminal output if you run the script from a terminal. This script writes the visualizations on the input frame and saves it as an image, after it creates a video with the frame images. You can check this code to make use of body_uv.

https://github.com/trrahul/densepose-video/blob/2157512c7604da1f00ef9a2065d4c81deff382df/vis.py#L384

[trrahul]

Also please note that this code was last updated 4 years ago

[arshikasoni]

This is running on detectron or detection2?