Integrating Mini 2 with CoreML

已完成

关注

Mike Oliver

• 2023年02月15日 01:38

I am developing an app that uses Mobile SDK to pull video from a Mini 2, run it through a Core ML model, and display an object classification. I am basing it off the sample DJI FPV app to pull video (https://riis.com/blog/dji-mobile-sdk-tutorials-ios-fpvdemo-swift/) and this iOS demo app (https://betterprogramming.pub/how-to-build-a-yolov5-object-detection-app-on-ios-39c8c77dfe58). It works but as soon as there is motion on the camera, the video blurs and no longer accurately provides an object classification. Is looks like a video buffer issue occurring when I call the CoreML model but I cannot get the issue to stop. I am using VideoProcesFrame and CreatePixelBuffer. Is there anything you can recommend?

0

• 

  DJI Developer Support

  • 2023年02月15日 03:56

  I would strongly suggest you switch to Android MSDK because the iOS MSDK project has been stopped. There is likely no more update in the future. Android MSDK 4.16.4 also supports the Mini 2 and it provides a higher stability in video stream.

  0

  评论操作 固定链接

请先登录再写评论。