Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

tflite model used in Google meet Virtual Background #1460

Closed
jiangjianping opened this issue Jan 3, 2021 · 12 comments
Closed

tflite model used in Google meet Virtual Background #1460

jiangjianping opened this issue Jan 3, 2021 · 12 comments
Assignees
Labels
type:feature Enhancement in the New Functionality or Request for a New Solution type:tflite TensorFlow Lite

Comments

@jiangjianping
Copy link

https://ai.googleblog.com/2020/10/background-features-in-google-meet.html

We are evaluating to add virtual background support to our app based on media pipe. Could you please open sourced the mobilenetv3 small tflite model used in Google Meet?

@Aayush795
Copy link

Aayush795 commented Jan 17, 2021

@jiangjianping Hey, I have downloaded the tflite file for Google Meets Virtual Background.
Do you think I can just use these models for my own application for the virtual background use case just like i use TF.js body pix models?

I am currently using Tensorflow JS models, BodyPix... but the accuracy is not that good. Can I somehow improve my model?

@floe
Copy link

floe commented Jan 18, 2021

According to the model card, the model is licensed under Apache License 2.0, so as long as your application has a compatible license, you should be good to go.

@Aayush795
Copy link

Aayush795 commented Jan 18, 2021

Thanks, @floe . However is it mandatory to convert the tflite model to tfjs model or I can use tf lite model as is? I want to use the model for real-time inference(Virtual Background/Blur) in a video web chat-based application that is utilizing Twilio video. Basically, I am looking for some open-source implementation for the same. something similar to what body pix already has
here
PS: I am new to this.

@floe
Copy link

floe commented Jan 18, 2021

Disclaimer: I am not a lawyer. That being said, I would assume that when your application source code is Apache 2.0 licensed, there shouldn't be any issues, no matter what format the model is in.

@fransiskusyoga
Copy link

@Aayush795 where can I find the model for the background segmentation?

@Aayush795
Copy link

@PINTO0309
Copy link

The license of MediaPipe Meet Segmentation Model seems to have been changed from Apache 2.0 to Google Terms of Service. I don't know if it is allowed to be used as open source, modified, or used commercially.

@floe
Copy link

floe commented Feb 5, 2021

Hmmm. Interesting. I wonder if they can change it retroactively, I would assume that only applies to newer releases?

@PINTO0309
Copy link

PINTO0309 commented Feb 5, 2021

I'm not a legal expert, so I can't say for sure, but I suspect it will be applied after the new release.

The following is just one Japanese lawyer's view.

==== Quote
Since the contract is concluded by accepting the terms proposed by the copyright holder at the time of licensing, if different license terms are proposed at that time, the license contract will be concluded based on the new terms. On the other hand, a contract that has already been concluded cannot be changed at the will of one party. Therefore, for software that has already started to be used, the license agreement has already been concluded under the license terms at that time, and those terms will continue to exist.
 :
 :
If you want to change the license terms to something else, all the copyright holders of the software in question must agree to change the license to that other license. Therefore, if there are countless programmers committed, as in the case of the Linux kernel, it would be very difficult to get everyone's consent, so changing the license would be virtually impossible.
==== Quote

@sgowroji sgowroji added type:tflite TensorFlow Lite stat:awaiting response Waiting for user response type:feature Enhancement in the New Functionality or Request for a New Solution labels Jul 20, 2021
@google-ml-butler
Copy link

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you.

@google-ml-butler
Copy link

Closing as stale. Please reopen if you'd like to work on this further.

@sgowroji sgowroji removed stale stat:awaiting response Waiting for user response labels Aug 3, 2021
@sgowroji sgowroji assigned sgowroji and unassigned mgyong Aug 3, 2021
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
type:feature Enhancement in the New Functionality or Request for a New Solution type:tflite TensorFlow Lite
Projects
None yet
Development

No branches or pull requests

8 participants