A Checklist for Training YOLOv3 for Our Own Dataset

a case study on object detection in image recognition

My last post “Exploring OpenCV’s Deep Learning Object Detection Library” had given a review on SSD/MobileNet and YOLOv2 under OpenCV 3.4.1 Deep Neural Network Module for object detection. It had also shown some examples detected by these two models. (ref: Figure 1 and Figure 2)

Figure 1.

Figure 2.

Because OpenCV 3.4.1 Deep Neural Network Module doesn’t support training on our own dataset, I am searching for other solutions which can support my future research on object detection. I have reviewed two implementations of YOLOv3 by Keras and Tensorflow on the Github:

• https://github.com/qqwweee/keras-yolo3
• https://github.com/experiencor/keras-yolo3

Here is a brief on the pre-trained models and the training steps under the implementation of qqwweee.

Detection by the Pre-trained Model

Figure 3. Tom Cruise in Mission Impossible 6.

Figure 4. Elephant, Giraffe, and Zebra.

Checklist for Training Your Dragon

(moving the scrollbar to check the full table)

What kind of object detection tasks I am working on? To be continued …

Acknowledgements: Thanks to the original work of YOLOv3 and the implementation by qqwweee.