Would you like to figure out how to demo your project?

[aimhabo]

I think I can imagine what cfg setting is. and whats the sample label rule do you set? is it a "binary image+src image" for every target? or some rule like "0 2532 12 2624 12 2881 10 3012 10 ..."

[ArtyZe]

Glad to tell you that the command in terminal is: ./darknet segmenter train ~.data ~.cfg ~.weights -gpus 0. Now you should know what is the image couple looks like (an RGB image and 0-1 image only)

[aimhabo]

unnn, maybe I haven't get it...

"Cannot load image "../darknet-mask/3c8bdfdb3.txt" "

I dont know why it loads image from txt. In my disk, darknet-mask/*.txt is the mask file (0 2532 12 2624 12 2881 10 3012 10 ...)

what should files be?

[ArtyZe]

the input_RGB_image and binary_label_image should be put in the same image, because in the data.c file, the "image load_image_gray(char *path, int w, int h)" function has been changed

[ArtyZe]

I use the cityscape dataset , for example, input_RGB_image(aachen_00001_000019_leftImg8bit.png) and the binary_label_image(aachen_00001_000019_gtFine_instanceIds.png) should be put in the same dir,

[hakS07]

@aimhabo Hi, I have recently started dealing with this repo and I don't get it how I can organize my data knowing that I have images and ground truth(masks) files?? what are darknet-labels(rectangles) and darknet-maks(seg.txt ) files??? how i can figure them out ??

can you help me

[aimhabo]

@essalahsouad just put RGB image and mask image in same file folder. darknet-labels is the detection-task label of bounding-box darknet-mask is the segment-task label of rle-mask Those things have no relationship with this project.

and about the rule of mask image, the author just updated the README and you can read it.

[hakS07]

@ArtyZe @aimhabo I followed the steps in the readme for the training 1)put label images(0-1) and original images(RGB) together:(_seg.bmp+_img.jpg)

2)modified in find_replace fonction ( find_replace(labelpath, "_img.jpg", "_seg.bmp", labelpath);) 3) generate the train.list file(path for the RGB image)

4)generate the .data file 5)modified in .cfg file 6) start train !./darknet segmenter train data/obj.data cfg/segment.cfg segment.backup -gpus 0. //////////But i got an error

cannot load image "/content/yolo_segmentation/project/C39_S1_I8_img.txt"
STB Reason:can't fopen

Any help, please

[chejulien]

Hi, I'am having the same problem as @essalahsouad and all my path setting was fine, did you guys figured out what's going on here?

[chejulien]

Problem solved, the root cause is find_replace in the image load_image_gray function was trying to change my path name...