yolov3 dataset formart

[bd7jat]

Hello, i fellow the instruction of https://github.com/qqwweee/keras-yolo3 and try to build custom dataset, One row for one image; Row format: image_file_path box1 box2 ... boxN; Box format: x_min,y_min,x_max,y_max,class_id (no space). and i found an example /home/zhangyang/yolo/VOC2007/JPEGImages/00040.jpg 64,52,162,106,1 /home/zhangyang/yolo/VOC2007/JPEGImages/00010.jpg 67,321,500,411,0 /home/zhangyang/yolo/VOC2007/JPEGImages/00007.jpg 36,342,451,483,0 So i use Yolo-Annotation-Tool-New to lable,and got lots txt file like this 0 0.508 0.545180722892 0.7 0.801204819277 from lable folder, they looks like not the same as keras-yolo3 required, I saw your post in keras-yolov3 issue264 , Please use my new yolo annotation tool. Because it will return the yolo format class_id,x_min,y_min,x_max,y_max not x,y,width and height. Did i do something wrong ?Or i need to do some extra job manully ?

[ManivannanMurugavel]

Hi @bd7jat , The correct yolo format is class_id,mid_x,mid_y,width,height

The all txt file has that corresponding image bbox values.