Ground Truth on predicted images

[Hishok]

@AlexeyAB @lukeai is there a way to show the ground truth on predicted images when using:

darknet.exe detector test data/obj.data yolo-obj.cfg yolo-obj_8000.weights

Thank you !

[johnny-mueller]

You can show the ground truth bounding boxes and also test the Data Augmentation on your dataset with this command:

./darknet detector -map -dont_show -show_imgs train cfg/coco.data cfg/yolov3.cfg

[Hishok]

Thank you @johnny-mueller , so is it only possible to show the ground truth on images when training the network using the -show_imgs flag?

[Sudhakar17]

@johnny-mueller @Hishok @AlexeyAB : How to print a class probability of predictions and GT box on the same image(save it) during testing?

[Tommy-gif-ai]

@AlexeyAB @LukeAI is there a way to show the ground truth on predicted images when using:

darknet.exe detector test data/obj.data yolo-obj.cfg yolo-obj_8000.weights

Thank you !

If I use this command (darknet.exe detector test data/obj.data yolo-obj.cfg yolo-obj_8000.weights),

can not see ground truth box, Can you teach to me? how can I see ground truth box on predicted images?

[stephanecharette]

I don't know if this would help you, but DarkMark has a toggle where it will show you the image markup, the predictions, or both together. The keyboard shortcut m toggles the marks, and p toggles the predictions. Examples:

Marks (annotations, ground truth):

Predictions:

Both at once:

[stephanecharette]

Otherwise, it should be very easy to write up a C++ application that shows both the predictions and the markup (when they exist) on all images in a directory. Come see us on the discord if you need help.

[matt-sharp]

@johnny-mueller @Hishok @AlexeyAB : How to print a class probability of predictions and GT box on the same image(save it) during testing?

@Sudhakar17 Did you manage to find a solution?

[Sudhakar17]

@matt-sharp Unfortunately not.