简体   繁体   中英

Poor face detection perfomance with Mobile Vision on Android platform

Good day to all.

So, as you can see from the title, the question is why is Mobile Vison face tracking works with so poor performance on Android platform compared to iOS?

I can get a smooth face tracking even on old iPhone 5s, but it's looks really ugly on Android devices even with Snapdragon 821 processors. I'm talking about Landmarks detection (eyes, ears, nose and mouth) with FaceTracker sample application.

I have tried to reduce Camera's pictureSize and previewSize to lower resolutions like 640x480 by rewriting CameraSource class, but the result is still bad and can't be compared to Apple devices.

Is this is Android platform problem at all, or it's because of Apple has really better hardware? Maybe someone knows how to optimize face detection with Mobile Vision on Android to get a better performance?

Thanks.

If you haven't touched the google sample code you might try running the code on other Android devices, as some CPUs may be optimized to work better than others even at different clock speeds. I recall experiencing reasonable performance on a Raspberry Pi.

If you have changed the code make sure you stay away from setBitmap in favor of setImageData when creating frames. Also, I'm not seeing face detection run at 30 FPS. If you are blocking frame rendering until detection returns, I'd expect choppy video at 1080p for sure, and likely lower res.

The technical post webpages of this site follow the CC BY-SA 4.0 protocol. If you need to reprint, please indicate the site URL or the original address.Any question please contact:yoyou2525@163.com.

 
粤ICP备18138465号  © 2020-2024 STACKOOM.COM