search

google-ai-edge / mediapipe

Cross-platform, customizable ML solutions for live and streaming media.
https://ai.google.dev/edge/mediapipe
Apache License 2.0
27.28k stars 5.13k forks source link

Missing Documentation for Compiling face_landmarker.task in MediaPipe Tasks Vision Package #5664

Open navimI opened 2 weeks ago

navimI commented 2 weeks ago

Description of issue (what needs changing)

Hi there,

I’ve been working with the MediaPipe Tasks Vision package and noticed that there is insufficient documentation on how to compile the face_landmarker.task file. This lack of information makes it challenging for new users and contributors to effectively utilize this functionality.

Clear description

Missing Documentation for Compiling face_landmarker.task in MediaPipe Tasks Vision Package

The expected documentation for the package shold be:

It would be even better for the newcommers to see some example of building this bundle.

Correct links

Task vision documentation

https://github.com/google-ai-edge/mediapipe/tree/master/mediapipe/tasks/web/vision

Provided pretrained model https://storage.googleapis.com/mediapipe-models/face_landmarker/face_landmarker/float16/1/face_landmarker.task

Parameters defined

No response

Returns defined

No response

Raises listed and defined

No response

Usage example

No response

Request visuals, if applicable

No response

Submit a pull request?

No response

ginesmartinezros commented 2 weeks ago

Same problem here. Any hint is appreciated.

kuaashish commented 2 weeks ago

Hi @navimI,

Thank you for raising this request. We agree that this should be documented as part of our process, as we have been receiving similar queries regarding converting customized and non-customized Tflite model to our supported Task format. We will bring this to our team’s attention, and if we have bandwidth’s, we will prioritize it. However, we are unable to provide a timeline for its availability at this moment.

Thank you!!

navimI commented 2 weeks ago

Hi @kuaashish,

Thank you for your response and for considering this request! I look forward to seeing detailed documentation for compiling the .task file. Additionally, I am planning to use the model with the BlazeFace (full-range) model for face detection. It would be helpful to have guidance on how to integrate this model as well, once the BlazeFace model is available.

Thanks again!