[RN] Pose Detection not working on android.
See original GitHub issueSystem information
- Have I written custom code (as opposed to using a stock example script provided in TensorFlow.js): Yes. I referred to the pose detection example.
- OS Platform and Distribution (e.g., Linux Ubuntu 16.04): MacOS Monterey 12.0.1
- Mobile device (e.g. iPhone 8, Pixel 2, Samsung Galaxy) if the issue happens on mobile device: Samsung Galaxy S10 5G
- TensorFlow.js installed from (npm or script link): npm
- TensorFlow.js version (use command below): ^0.8.0
- Browser version: Chrome 96.0
- Tensorflow.js Converter Version: ^3.11.0
- Expo version : 43.0.0
Describe the current behavior Pose estimation not working properly only on Android. (keypoints are detected, but very low score)
There was no error during preparation (load model, handleImageTensorReady). Also, camera preview works properly.
Describe the expected behavior Pose estimation should work same as IOS.
Standalone code to reproduce the issue Provide a reproducible test case that is the bare minimum necessary to generate the problem. If possible, please share a link to Colab/CodePen/any notebook.
Other info / logs Include any logs or source code that would be helpful to diagnose the problem. If including tracebacks, please include the full traceback. Large logs and files should be attached.
package.json
"@tensorflow-models/pose-detection": "^0.0.6",
"@tensorflow/tfjs": "^3.11.0",
"@tensorflow/tfjs-react-native": "^0.8.0",
"expo": "~43.0.0",
"expo-camera": "~12.0.3",
"expo-file-system": "~13.0.3",
"expo-gl": "~11.0.3",
"expo-gl-cpp": "~11.0.1",
Issue Analytics
- State:
- Created 2 years ago
- Comments:9
Top Results From Across the Web
Detect poses with ML Kit on Android | Google Developers
The pose detector will first detect the most prominent person in the image and then run pose detection. In subsequent frames, the person-detection...
Read more >Learning Human Pose Estimation Features with Convolutional ...
This paper introduces a new architecture for human pose estimation using a multi- layer convolutional network architecture and a modified ...
Read more >Google MLKit pose detection not executing ...
I am using MediaMetadataRetriever as well as MLkit to do pose detection on a saved video file. I have attached some of my...
Read more >3D Pose Detection of Closely Interactive Humans Using Multi ...
To solve this problem, we first obtain 2D joints in every image using OpenPose and human semantic segmentation results from Mask R-CNN. With...
Read more >How TikTok Became a Diplomatic Crisis - The New York Times
Born in 1983 as the only son of a librarian and a nurse, Zhang came of age in a ... The problem with...
Read more >Top Related Medium Post
No results found
Top Related StackOverflow Question
No results found
Troubleshoot Live Code
Lightrun enables developers to add logs, metrics and snapshots to live code - no restarts or redeploys required.
Start FreeTop Related Reddit Thread
No results found
Top Related Hackernoon Post
No results found
Top Related Tweet
No results found
Top Related Dev.to Post
No results found
Top Related Hashnode Post
No results found
Top GitHub Comments
@rthadur I tested the working example on both OS, but neither worked. (Galaxy S10 5G, iPhone 11 Pro)
Result IOS : app crashed after
warning: initialization of backend rn-webgl failed
Android : only black screen displayed.(both front/back camera) frame rate was about 4~5fps.Test Process
yarn && yarn start
@rthadur @jinjingforever Thank you for the update. I’ve tested the example on Samsung Galaxy S10, but the example still not working properly.
Expected behavior keypoints should be displayed just like the image on the
README.md
Actual behavior App does not crashed, but the keypoints(green circles) not displayed properly. (only 1-2 circles shown randomly)
Test Process
yarn
=> Failed witherror An unexpected error occurred: "https://registry.yarnpkg.com/colors/-/colors-1.4.2.tgz: Request failed \"404 Not Found\"".
rm yarn.lock
yarn && yarn start
README.md
)