Darknet: Coordinations of bounding boxes

[mfaramarzi]

@AlexeyAB
Hello dear Alex
How can I modify the "image.c" to only get four numbers for the coordinates of bounding boxes (without any additional description)?
I have tried by making some minor changes as below (and running "make" command) but it didn't make any change to the output as could be seen in the second image:

[stephanecharette]

Instead of modifying Darknet itself, have you thought about using the darknet library to have your own code display exactly what you need/want?

Somewhat related to that, you may be interested in knowing there are other 3rd-party libraries that use libdarknet.so which can output the results. For example, I wrote a free one called DarkHelp which has an option to output the coordinates as a JSON file. Here is an example of the output:

[...]
                    "rect": {
                        "height": 141,
                        "width": 171,
                        "x": 44,
                        "y": 26

See here for details: https://www.ccoderun.ca/darkhelp/api/ShellScripting.html

[mfaramarzi]

Instead of modifying Darknet itself, have you thought about using the darknet library to have your own code display exactly what you need/want?
Somewhat related to that, you may be interested in knowing there are other 3rd-party libraries that use libdarknet.so which can output the results. For example, I wrote a free one called DarkHelp which has an option to output the coordinates as a JSON file. Here is an example of the output:

[...]
                    "rect": {
                        "height": 141,
                        "width": 171,
                        "x": 44,
                        "y": 26

See here for details: https://www.ccoderun.ca/darkhelp/api/ShellScripting.html

I just got familiar with "libdarknet.so" and trying to figure out how to use it for my purpose.
Meanwhile, as you can see in the "result.txt" some of objects have negative value coordination and large width/height values (e.g. "2.jpg" --> pavement: 96% (left_x: -39 top_y: -4 width: 548 height: 660) Enter Image Path: data/obj/3.jpg: Predicted in 9.295000 milli-seconds.),

while according to the here:

struct bbox_t { unsigned int x, y, w, h; // (x,y) - top-left corner, (w, h) - width & height of bounded box

Q-1: How can I interpret negative values of x and y?
Q-2: How can I interpret width/height larger than 416, while I I used 416*416 in the .cfg file?

[stephanecharette]

Q-1: How can I interpret negative values of x and y?
Q-2: How can I interpret width/height larger than 416, while I I used 416*416 in the .cfg file?

Note that in DarkHelp, this is addressed with DarkHelp::fix_out_of_bound_values, which is enabled by default.

[mfaramarzi]

Q-1: How can I interpret negative values of x and y?
Q-2: How can I interpret width/height larger than 416, while I I used 416*416 in the .cfg file?

Note that in DarkHelp, this is addressed with DarkHelp::fix_out_of_bound_values, which is enabled by default.

That would be great. But unfortunately I have no experience with C++ and want to embed my model into a Python application.

[mfaramarzi]

Instead of modifying Darknet itself, have you thought about using the darknet library to have your own code display exactly what you need/want?

Somewhat related to that, you may be interested in knowing there are other 3rd-party libraries that use libdarknet.so which can output the results. For example, I wrote a free one called DarkHelp which has an option to output the coordinates as a JSON file. Here is an example of the output:

[...]
                    "rect": {
                        "height": 141,
                        "width": 171,
                        "x": 44,
                        "y": 26

See here for details: https://www.ccoderun.ca/darkhelp/api/ShellScripting.html

I just got familiar with "libdarknet.so" and trying to figure out how to use it for my purpose.
Meanwhile, as you can see in the "result.txt" some of objects have negative value coordination and large width/height values (e.g. "2.jpg" --> pavement: 96% (left_x: -39 top_y: -4 width: 548 height: 660) Enter Image Path: data/obj/3.jpg: Predicted in 9.295000 milli-seconds.),

while according to the here:

struct bbox_t { unsigned int x, y, w, h; // (x,y) - top-left corner, (w, h) - width & height of bounded box

Q-1: How can I interpret negative values of x and y?
Q-2: How can I interpret width/height larger than 416, while I I used 416*416 in the .cfg file?

@cenit
Would you please help me with those two questions? Tanks in advance!