Real-Time Object Detection
Faster Real-Time Object Detection: YoloV4 in Pytorch
How I am detecting my lovely cat faster than ever!
It’s been multiple years since the groundbreaking You Only Look Once first appeared. It went through 3 versions, respectively Yolo, YoloV2, YoloV3. The third one which until very recently was still considered to be state of the art (even though sort of replaced by EfficientDet, FasterRCNN methods for instance) in the Yolo architecture trilogy has been topped by a newer version, the YoloV4, on 23rd April 2020. For me personally, it is a true matter of rejoicing, I have been using Yolo architecture for a moment now, and always ended up hitting the wall of Real-Time Inference on my poor RTX 2060 (not that bad actually) that I was borrowing during my job at a Tech Company. So just like any other matter of interest, I will be giving a short review of the paper for YoloV4 (you can all links at the end of the article) and showing you its implementation in PyTorch plus a glimpse of its inferences with source code added at the end!
Now let’s get started!
