Yolov3 Object Detection - Data Sets

Hello,

I am attempting to use yolov3 object detection algorithm for an automous vehicle project for a senior design course.

I would like to train the network to detect more objects; like stairs. I would also like to eliminate alot of unuseful classes from the coco data set, like a hair brush as the automous vehicle will never encounter that sort of thing.

Are there any matlab resources on how to label images and import them into the matlab workspace for yolov3 training?

I am unable to find relevant resources for this task.

Any tips would be helpful,

Thank you.

Answer 1

[Srivardhan Gadila]

For labelling the images, you can use the Image Labeler app. Refer to the following resources related to labelling and preparing the training data:

1. Get Started with the Image Labeler
2. Training Data for Object Detection and Semantic Segmentation

For in general transfer learning worfklow, refer to the following: Transfer Learning in MATLAB.

Refer to the following for information related to yolov3 network in MATLAB for object detection, transfer learning etc.

1. Getting Started with YOLO v3
2. yolov3ObjectDetector
3. Object Detection