taoo0316 / trying-out-yolo

trying-out-YOLO

Trying out YOLO for real-time object detection

https://pjreddie.com/darknet/yolo/

Successful Classification of .JPEG Image

Multiple Images

Tiny YOLO

There were fewer layers in this case, and the prediction took significantly less time.

Open Images dataset

We could also use the Open Images dataset. In this case, the predictions were less precise and took longer to complete.

Training YOLO on COCO

I have also set up the necessay requirements to train YOLO on COCO on my local machine. Unfortunately it takes too long to train on a MacBook. Nonetheless, it was a good exercise.

Reflections

Overall, it was interesting to try out the different functionalities of YOLO and try to train the model on my local machine. Through the process, I also got a better understanding of the trade-off between computational efficiency and accuracy.

I have also began reading the paper published by Joseph Redmon to learn more about the theoretical foundation of YOLO. The pdf can be found in this repository.

The darknet folder is too large in size to view. Nonetheless, let me know if you want to view the code base as well.