https://github.com/opendatacam/demo-object-detection-browser
A technical prototype to test the posibility of runinng object detection in the browser of your phone or computer
https://github.com/opendatacam/demo-object-detection-browser
camera coco-ssd computer-vision cv mobile object-detection object-tracking tensorflowjs
Last synced: 4 months ago
JSON representation
A technical prototype to test the posibility of runinng object detection in the browser of your phone or computer
- Host: GitHub
- URL: https://github.com/opendatacam/demo-object-detection-browser
- Owner: opendatacam
- License: mit
- Created: 2022-04-03T21:19:06.000Z (over 4 years ago)
- Default Branch: main
- Last Pushed: 2023-08-18T14:21:28.000Z (almost 3 years ago)
- Last Synced: 2025-02-27T11:39:36.465Z (over 1 year ago)
- Topics: camera, coco-ssd, computer-vision, cv, mobile, object-detection, object-tracking, tensorflowjs
- Language: HTML
- Homepage: https://opendatacam.github.io/demo-object-detection-browser/
- Size: 49.8 KB
- Stars: 1
- Watchers: 1
- Forks: 1
- Open Issues: 2
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Object Detection in the Browser
This is a small benchmark to see how fast the [coco-ssd](https://github.com/tensorflow/tfjs-models/tree/master/coco-ssd) runs on TensorFlow.js in the browser.
To run the benchmark visit https://opendatacam.github.io/demo-object-detection-browser/.
## Bechmark Results
| Device | Average FPS | Poor Chart™️
(= ~ 5 fps) |
|----|----|----|
| iPhone 13 (Safari) | 20 | ==== |
| Samsung Galaxy S21 (Chrome) | 14 - 18 | ==== |
| iPhone 11 (iOS 15.4.1, Safari) | 15 - 18 | ==== |
| iPhone 12 64 GB (iOS 15.1, Safari) | 12 - 14 | === |
| iPhone 12 64 GB (iOS 15.1, Chrome) | 12 - 14 | === |
| iPhone 12 Pro Max | 12 - 14 | === |
| iPad Mini 6 (Safari) | 14 | === |
| Pixel 4 | 11 | === |
| iPad Pro 10.5" (iOS 15.4.1, Safari) | 9 | == |
| Xiaomi Mi 8 | 5 - 7 | == |
| Google Pixel 6a (Android 13, Chrome) | 5 | = |
| iPhone 6S (Safari) | 5 | = |
| Samsung Galaxy Tab A T510 | 2 - 4 | |
| | | |
| MacBook Pro 14 M1 Pro (Chrome) | 60 | ============ |
| MacBook Pro 13 M1 (Chrome) | 40 | ======== |
| MacBook Pro 14 M1 Pro (Safari) | 30 | ====== |
| MacBook Pro 13 M1 (Safari) | 14 - 20 | ==== |
| MacBook Pro 13 (2018, 2.7 GHz i7, Intel Iris Plus Graphics 655 1536 MB, Firefox) | 18 - 20 | ==== |
## Contribute
Please feel free to run the demo on your devices and upload your results to GitHub (via PR or as Issues) or send me a message via Mastodon: https://mastodon.social/@vsaw
## Known Issues
- Detection Boxes may have the wrong position for some devices (e.g. iPhone 11 in Portrait mode). As a simple workaround I suggest you try to change the orientation of your device if possible (e.g. use in Landscape mode).
## See also
- https://www.tensorflow.org/js
- https://github.com/tensorflow/tfjs-models/tree/master/coco-ssd
- https://github.com/nihui/ncnn-webassembly-nanodet/ another benchmark running ncnn on Web Assembly