구글의 object detection 모델 MoveNet 관련 자료

2022. 5. 9. 17:19서버 프로그래밍

https://blog.tensorflow.org/2021/05/next-generation-pose-detection-with-movenet-and-tensorflowjs.html

 

Next-Generation Pose Detection with MoveNet and TensorFlow.js

MoveNet is a human pose detection architecture developed at Google that is ultra fast and accurate. It was designed to detect difficult poses

blog.tensorflow.org

- js기반의 텐서플로우로 설계되어 server call 없이 오직 client 단에서도 동작 가능하여 고객의 개인 디바이스(데스크탑, 랩탑, 스마트폰)에서 실행할 수 있는 환경을 고려한 모델
- 모델 사이즈는 크지 않음 (20MB 전후)
- 그동안은 모델의 성능과 정확도를 높이는데 주력해왔고 실제로 최적화하여 앱, 웹 등에 제품화 되기 시작한건 얼마 되지 않음 (거의 1년 정도)

https://blog.tensorflow.org/2021/08/pose-estimation-and-classification-on-edge-devices-with-MoveNet-and-TensorFlow-Lite.html

 

Pose estimation and classification on edge devices with MoveNet and TensorFlow Lite

MoveNet is the state-of-the-art pose estimation model for mobile devices which can run in realtime on modern smartphones. This blog post teaches you

blog.tensorflow.org

 

안드로이드, 아이폰 예제도 친절하게 제공해주고 있다.

https://github.com/tensorflow/examples/tree/master/lite/examples/pose_estimation/android

 

GitHub - tensorflow/examples: TensorFlow examples

TensorFlow examples. Contribute to tensorflow/examples development by creating an account on GitHub.

github.com

https://github.com/tensorflow/examples/tree/master/lite/examples/pose_estimation/ios

 

GitHub - tensorflow/examples: TensorFlow examples

TensorFlow examples. Contribute to tensorflow/examples development by creating an account on GitHub.

github.com

 

기타 참고 블로그

https://inhopp.github.io/tf_tutorial/Tutorial1/

 

[TF_Tutorial] MoveNet - Pose Detection

Main Reference: - TensorFlow Blog : MoveNet - An Overview of Human Pose Estimation with Deep Learning

inhopp.github.io