AlexeyAB / darknet

YOLOv4 / Scaled-YOLOv4 / YOLO - Neural Networks for Object Detection (Windows and Linux version of Darknet )
http://pjreddie.com/darknet/
Other
21.61k stars 7.95k forks source link

Yolov4-tiny_3l for small and large object? #6210

Open nguyenanhtuan1008 opened 4 years ago

nguyenanhtuan1008 commented 4 years ago

I would like to train the model on Yolov4 tiny with small and large objects. But I found in respo only Yolov3-tiny_3l.cfg: https://raw.githubusercontent.com/AlexeyAB/darknet/master/cfg/yolov3-tiny_3l.cfg

@AlexeyAB Can you update the config file for this?

Thank you so much!

AlexeyAB commented 4 years ago

I added these models: https://github.com/AlexeyAB/darknet/commit/de68e19cc627f642023f09513ac2306fbcbc1e4b

nguyenanhtuan1008 commented 4 years ago

Thank you so much for quick reply. @AlexeyAB 🙏.

nguyenanhtuan1008 commented 4 years ago

Sorry I have another question about anchor calculation in this yolov3-tiny_3l.cfg: Which dataset did you use for anchors calculation? COCO dataset? anchors = 12, 16, 19, 36, 40, 28, 36, 75, 76, 55, 72, 146, 142, 110, 192, 243, 459, 401

I have try to do it with my custom dataset by command: darknet.exe detector calc_anchors data/obj.data -num_of_clusters 9 -width 416 -height 416

But when I training the model the mAP seen less than your default anchors around 1%.