CODEWITHSUNDEEP
deep-learningcomputer-visionobject-detectionyolov5video-tracking
Muhammad Daniyal
Muhammad Daniyal

Reputation: 23

Object tracking of fast moving small objects

I have a task to count small objects falling freely from conveyor exit using camera. Camera input will be like as shown in this video below, and I need to count them.

https://drive.google.com/file/d/1nKzDVLBZmpdQDIsiNU1BwiwKKPdm-dHo/view

I was thinking to go for object detection using Yolo and then apply DeepSort for tracking and counting.

I have two simple questions regarding this.

  1. Is it good to use Yolo detection for such fast moving object problem or is it a bad approach. And if it's not right, what approach would you suggest me?
  2. Secondly, I don't have image dataset available, I only have these below videos, which I'm thinking to extract frames from and label them for object detection. But the issue is if you check the video, the video quality is quite poor so the frames from video doesn't show objects clearly for annotations. (https://drive.google.com/file/d/1yCNu-1ViXYlqWBlKqzxWtXBueYXgmlXS/view?usp=share_link) (https://drive.google.com/file/d/1nKzDVLBZmpdQDIsiNU1BwiwKKPdm-dHo/view?usp=share_link)

Is there any solution to above problems. And what should be the approach and methods to count such fast moving small objects?

Any help of any kind will be very appreciated :)

Upvotes: 0

Views: 1110

Answers (1)

Mike B
Mike B

Reputation: 3476

When you pause the video on any of the frames you all the moving objects are smeared out on the image. Your camera is your main issue here. My suggestion is that you first get a camera with a higher shutter speed so that the moving objects appear sharper. Then, collect a dataset, annotate and train Yolo.

Feel free to checkout out my tracking repo: https://github.com/mikel-brostrom/Yolov5_StrongSORT_OSNet

Upvotes: 0

Related Questions