How to make my own YOLO-pose data set?（JSON2YOLO）

[aoqiangma]

Hello, everyone, I just started learning YOLO-pose algorithm. The data set format of our YOLO-pose algorithm is YOLO format, but labelme tagging software gets JSON format. Excuse me, everyone: Is there any code that can convert JSON format to YOLO format? By the way, all the conversion codes I found are rectangular frame coordinates, not key point coordinates. I hope you can help me, thank you very much.

[vaibhav030798]

can you provide me this scripts i am not understand this process thankyou

[Sridharanraja]

I don't understand this scripts, therefore could you perhaps provide me a brief explanation?

[vaibhav030798]

and the places where 17 and 51 appear also need to be changed, the original is 17 The key point, 17*3=51, only need to change these?

[vaibhav030798]

after apply this
my result is

0 0.0 0.0 0.0 0.0 1 0.0 0.0 0.0 0.0 2 0.0 0.0 0.0 0.0 3 0.0 0.0 0.0 0.0 4 0.635156 0.731944 0.260937 0.402778

but i need

0 0.307292 0.497656 0.352083 0.782813 0.15625 0.1375 2 0.172917 0.820312 2 0.45 0.834375 2 0.489583 0.151562 2

[vaibhav030798]

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 8/8 [00:00<00:00, 11606.51it/s] 0%| | 0/40 [00:00<?, ?it/s] Traceback (most recent call last): File "final.py", line 40, in img_width = annotations["width"] KeyError: 'width'

[vaibhav030798]

please provide me information its urjent

[vaibhav030798]

i also mark it with coco-annotator

[vaibhav030798]

i clear my problem

my landmark point is 4 and one bounding box cord which is annotate help of coco annotator and trhis coco annotator json file convert to yolov7 txt file format

[vaibhav030798]

dog_face_new (copy).txt

this is json file

[Sridharanraja]

ep-1.zip I can convert json to txt format for bbox using this as my source file, but I also want to convert keypoint; the second code I tried to use to do so didn't work, and I'm using coco-annotator for this annotation.

[vaibhav030798]

i also face this problem

[Sridharanraja]

i got these valuse,

3 0.716797 0.553472 0.088281 0.184722 0 0.0 0.0 0.0 0.0 1 0.0 0.0 0.0 0.0 2 0.0 0.0 0.0 0.0

[vaibhav030798]

same but need

0 0.307292 0.497656 0.352083 0.782813 0.15625 0.1375 2 0.172917 0.820312 2 0.45 0.834375 2 0.489583 0.151562 2

[Sridharanraja]

but you got it or not?

[vaibhav030798]

not

[Sridharanraja]

you convert keypoint also or not because why i asking question i have been convert only boundingbox not keypoints like for example, 3 0.716797 0.553472 0.088281 0.184722 -->It's classes and bbox value 0 0.0 0.0 0.0 0.0 -->KP 1 0.0 0.0 0.0 0.0 -->KP 2 0.0 0.0 0.0 0.0 -->KP

[Sridharanraja]

@Sridharanraja I sent a test.zip, after decompression, just run trans.ty, if you use the json file obtained by coco-annotator, rename it to train.json, replace where I am, and replace 11 and 33 with Yours, 11 is the key point, 33 is three times of 11

@lengdanlexin yeah i tried this method, but i faced same thing like this👇

[Sridharanraja]

@lengdanlexin I am aware of the codes mentioned above, but there are only three keyholes in this situation.

[Madhan-Kumarr-D]

where is the python sript to convert json format annotations into YOLO format

[lengdanlexin]

I re-wrote a tool for marking key points by myself, only need to install opencv-python.All details are at the URL below

https://blog.csdn.net/qq_45149610/article/details/130757471?csdn_share_tail=%7B%22type%22%3A%22blog%22%2C%22rType%22%3A%22article%22%2C%22rId%22%3A%22130757471%22%2C%22source%22%3A%22qq_45149610%22%7D

The code is also placed in this txt file. There are only two places where the key points need to be changed, keypoint.txt Set the corresponding key_point_num # number of key points Replace the path where the image is located and the save path, and nothing can be placed in the save path. If the label already exists locally, if the prefix and suffix are the same, the local folder will be loaded onto the image first. Run the program directly, and the local label can be loaded correctly. Like ordinary labels, a green hollow circle will appear around the 8 points of the mouse and the mouse near the key points. When the hollow circle appears, you can drag the key points and boxes. The difference from ordinary annotations is that double-clicking the rectangle here will not cause any operation, which means that you can't delete the rectangle, you can only move it, and you can't delete the key point, you can only move it. Double-click the key point with the left mouse button, and enter 0 or 1 and 2 from the keyboard to switch whether the key point is visible. If there is no local label, a box and key_point_num key points will be generated, and the corresponding label will be generated locally. When the map is marked, click Q to switch, and you can see that the corresponding key point labels are generated locally. For more details, please refer to the blog. @vaibhav030798 @Sridharanraja @Madhan-Kumarr-D @aoqiangma @taka-wang

[MrFahad]

I have TWO questions:

1 - Iis it possible to have different keypoints for different classes?? If yes can anyone tell me how to deal with that?? Like I have generated the .json file along with the bbox and keypoints for each object in the image.

2 - One more thing, if one image have multiple objects and they are from different classes, then its obvious that their keypoints are also different. How to deal with that scenario as well.

Please guide me in this regard, Thanks in advance.