Way to save a frame of outputvideo with the corresponding .txt-file?

[Motch93]

I run the yolov3 as videodetection on my webcam and i want to save a jpg or png frame with the corresponding .txt file (class,coordinates). So i get the images with .txt-files while running the code. Is there already a solution or how do i have to modificate the code? Thanks for your help. (Yolov3 on Jetsontx2 ubuntu with GPU=1, CUDNN=1)

[Zuo1ji]

do you mean you want to know where to find the source code of the img?

if so,

there is a function in src/image.c called 'show_image_cv' may help you know where the img is，Suppose you have used opencv

if you donot have opencv, there is another function in src/image.c called ‘save_image_png’ shows where the data is

[AlexeyAB]

@Motch93

Do you get test.mp4 and do you want to get frame1.jpg, frame2.jpg,,, with correspond txt files for each jpg-image?

• You can try to use: https://github.com/AlexeyAB/Yolo_mark to get frames mkdir img ./yolo_mark /img cap_video test.mp4 1 so you will get all frames (jpg-images) in the /img directory

Then run command - to get list of all images in the train.txt file: ./yolo_mark /img train.txt coco.names

• Then you can run the Darknet by using this command: ./darknet detector test cfg/coco.data yolov3.cfg yolov3.weights -dont_show -ext_output -save_labels < data/train.txt

As result you will get txt-label files in the /img directory for each jpg-file. But you will get it in Yolo format: <class> <x_relative_center> <y_relative_center> <width_relative_center> <height_relative_center>