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
27.49k stars 5.15k forks source link

How to build face_landmarks.tflite from mediapipe graph? #4454

Closed nguyennyssf closed 1 year ago

nguyennyssf commented 1 year ago

OS Platform and Distribution

macOS Ventura 13.2.1

Compiler version

No response

Programming Language and version

TFLite

Installed using virtualenv? pip? Conda?(if python)

No response

MediaPipe version

Newest from source with branch master

Bazel version

5.2.0

XCode and Tulsi versions (if iOS)

No response

Android SDK and NDK versions (if android)

No response

Android AAR (if android)

None

OpenCV version (if running on desktop)

No response

Describe the problem

I just want to build model .tflite from the graph of face landmark.
Please help me or give me the instruction how to do that. Really appreciate for your helps

Complete Logs

(Don't know how to build .tflite from .pbtxt of face_landmark)
ayushgdev commented 1 year ago

Hello @nguyennyssf TfLite models are artifacts which cannot be created from MediaPipe graphs. They can only be used in MediaPipe graph like here in face landmarks. These can be downloaded from the Model cards page/documentation page of the corresponding solution. For face landmark, the model can be found in this section.

nguyencse commented 1 year ago

Yes, thanks for your response @ayushgdev. In fact that I really really need to build an ios framework use the latest model of facemesh. I cannot build it on macOS Ventura with XCode 14.3 (macOS Ventura only supports XCode 14 and so I cannot downgrade version). Really need help or could you give me an example of facemesh xcodeproj? I cannot install Tusi app as well due to macOS version.

ayushgdev commented 1 year ago

@nguyencse We are using MacOS Ventura 13.4 with Xcode 14.1. You can build the legacy solutions in Tulsi with this version setup. However, iOS APIs are getting major upgrades since we are ending support for these MediaPipe Legacy Solutions, and upgrading the others. We are yet to release the iOS MediaPipe Task APIs and we would request you to wait for sometime. These APIs will be much simpler to use.

nguyencse commented 1 year ago

ok thanks @ayushgdev , I built a temp framework of my own in ios and waiting for your task api. Thank you

google-ml-butler[bot] commented 1 year ago

Are you satisfied with the resolution of your issue? Yes No