search

AlexeyAB / darknet

YOLOv4 / Scaled-YOLOv4 / YOLO - Neural Networks for Object Detection (Windows and Linux version of Darknet )
http://pjreddie.com/darknet/
Other
21.8k stars 7.97k forks source link

Using YOLO v2 / v3 with single anchor box #1513

Open groot-1313 opened 6 years ago

groot-1313 commented 6 years ago

Hi, I would like to know if it is possible to use YOLO v2 / v3 with a single anchor box per cell because our application requires only one object detection per cell (there is only 1 kind of object split into 2 classes) and we are looking at all possible methods to reduce the architecture. So far, we have used YOLO v1 and are satisfied with the results, but as mentioned in by @AlexeyAB in this issue, YOLO v1 is not very accurate, so we want to improve the performance and explore v2 and v3. Any help will be very much appreciated.

AlexeyAB commented 6 years ago

@groot-1313

I would like to know if it is possible to use YOLO v2 / v3 with a single anchor box per cell because our application requires only one object detection per cell

Yes, it is possible. Try to train yolov3-tiny.cfg as described here: https://github.com/AlexeyAB/darknet#how-to-train-tiny-yolo-to-detect-your-custom-objects

Add here your 1 anchors (2 values):

Set num=1

Also