Detection and network problems

AlexeyAB / darknet

YOLOv4 / Scaled-YOLOv4 / YOLO - Neural Networks for Object Detection (Windows and Linux version of Darknet )

http://pjreddie.com/darknet/

Other

21.75k stars 7.96k forks source link

Detection and network problems #2848

Open chi8411 opened 5 years ago

chi8411 commented 5 years ago

Hi,I want detection traffic sign from big images(2048*2048). What can I do to improve the accuracy? And I want to try to change the network. Should I just change the convolution of cfg? If I want to reduce the number of downsampling layers in Yolov3, how can I do? If I want to pre-process the image, where should I add it? It only detects the top half of the image or 10% under the image is ignored. Thank you.

AlexeyAB commented 5 years ago

@chi8411 Hi,

You should use width=832 height=832 or width=1024 height=1024 in your cfg-file
Also you can try to use these modified cfg-files:
- https://github.com/AlexeyAB/darknet/blob/master/cfg/yolov3_5l.cfg
- https://github.com/AlexeyAB/darknet/blob/master/cfg/yolov3-tiny_3l.cfg

Show anchors and cloud.png images that will be generated by using command: https://github.com/AlexeyAB/darknet#how-to-improve-object-detection ./darknet detector calc_anchors data/obj.data -num_of_clusters 9 -width 832 -height 832 -show

chi8411 commented 5 years ago

@AlexeyAB Hi,I have a problem about "anchors box" yolov3 need 9 anchors . This is my calculation. anchors = 7, 8, 11, 12, 16, 18, 24, 25, 21, 35, 33, 35, 33, 50, 46, 65, 66, 96 And yolov3_5l need 15 anchors.Is it? This is my calculation. anchors = 10, 11, 13, 14, 18, 19, 22, 24, 27, 30, 34, 35, 31, 53, 42, 43, 52, 54, 43, 70, 53, 87, 68, 69, 72,100, 87,127, 120,174

Is this calculation correct? I used "./darknet detector calc_anchors ../../new/traffic_sign.data -num_of_clusters 15 -width 1024 -height 1024"

Thank you.

AlexeyAB commented 5 years ago

Is this calculation correct? I used "./darknet detector calc_anchors ../../new/traffic_sign.data -num_of_clusters 15 -width 1024 -height 1024"

Yes.