Yolov5: How do i generate result.json file?

[SamiurRahman1]

❔Question

I want to run yolo on a bunch of images that i got from a video and i want to save the results as result.json file. The only place i found something about .json file is in test.py. But how do i use it on my images?

Additional context

[github-actions[bot]]

Hello @SamiurRahman1, thank you for your interest in our work! Please visit our Custom Training Tutorial to get started, and see our Jupyter Notebook , Docker Image, and Google Cloud Quickstart Guide for example environments.

If this is a bug report, please provide screenshots and minimum viable code to reproduce your issue, otherwise we can not help you.

If this is a custom model or data training question, please note that Ultralytics does not provide free personal support. As a leader in vision ML and AI, we do offer professional consulting, from simple expert advice up to delivery of fully customized, end-to-end production solutions for our clients, such as:

• Cloud-based AI systems operating on hundreds of HD video streams in realtime.
• Edge AI integrated into custom iOS and Android apps for realtime 30 FPS video inference.
• Custom data training, hyperparameter evolution, and model exportation to any destination.

For more information please visit https://www.ultralytics.com.

[glenn-jocher]

@SamiurRahman1 you can create a coco style json file using test.py:
python test.py --save-json

You can save text labels using detect.py:
python detect.py --save-txt

[SamiurRahman1]

@SamiurRahman1 you can create a coco style json file using test.py:
python test.py --save-json

You can save text labels using detect.py:
python detect.py --save-txt

Thank you for your reply. But if i am using test.py, how do i give the input image files to the script?

[glenn-jocher]

@SamiurRahman1 the same way it's currently used.

[SamiurRahman1]

@SamiurRahman1 the same way it's currently used.

Would you mind elaborating a bit? i am very new to yolo. I ran the following code:
python test.py --save-json --data /home/sam/Downloads/processed_images2/2/300_4_3.jpg

and i get this error:
UnicodeDecodeError: 'utf-8' codec can't decode byte 0xff in position 0: invalid start byte

[glenn-jocher]

@SamiurRahman1 start from the tutorials https://github.com/ultralytics/yolov5#tutorials

[SamiurRahman1]

So, as much as i understand, i need a .yaml file of my dataset to use test.py. For that, i need to train using my dataset. Am i on the right track?

[glenn-jocher]

@SamiurRahman1 yes you want a data.yaml file that shows training and testing image folders. Then you just point test.py to it, it will evaluate all val images.
https://github.com/ultralytics/yolov5/blob/90e7bfdf7561b52433ae2974445c4eaec22ad536/data/coco128.yaml#L10-L13

python test.py --data coco128.yaml

[SamiurRahman1]

@SamiurRahman1 yes you want a data.yaml file that shows training and testing image folders. Then you just point test.py to it, it will evaluate all val images.

https://github.com/ultralytics/yolov5/blob/90e7bfdf7561b52433ae2974445c4eaec22ad536/data/coco128.yaml#L10-L13

python test.py --data coco128.yaml

Ok thank you. I got it now. But what if i don't want to train, just test my dataset with a pre-trained model and want to save the results as .json file? is there anyway to do that?

[Denizzje]

+1 for this if it not possible right now. With Darknet Yolo, I use the generate result.json to analyse the detections further. The save-txt argument seems to just make labels for every picture in a seperate file. I would like to have all detections in one json (or something comparable) file.

[glenn-jocher]

@SamiurRahman1 yes of course it's possible. You just follow the steps I already explained to you.

[glenn-jocher]

@Denizzje follow the directions I explained already above.

[SamiurRahman1]

@SamiurRahman1 yes of course it's possible. You just follow the steps I already explained to you.

when i try to run the script without training, it asks for a cache file, which i assume i only get when i train?

[glenn-jocher]

@SamiurRahman1 no

[SamiurRahman1]

ok.. never mind. i got my json file using yolov3. this wasn't helpful for me and i think it is better if i leave it at that. thank you for your replies.

[SamiurRahman1]

FYI, i changed the yaml file with where my image files are:

# train and val data as 1) directory: path/images/, 2) file: path/images.txt, or 3) list: [path1/images/, path2/images/]
train: /home/sam/Downloads/SIT/darknet-master/data/list1.txt  # 128 images
val: /home/sam/Downloads/SIT/darknet-master/data/list1.txt  # 128 images

The file is surely there and the content of the file is full path for my images.

/home/sam/Downloads/SIT/KevalDoshi/ori_images/2/10000.jpg 

/home/sam/Downloads/SIT/KevalDoshi/ori_images/2/1000.jpg

and so on. This is the error i get as i mentioned. It is looking for a .cache file.

line 215, in __init__
    super(_open_file, self).__init__(open(name, mode))
FileNotFoundError: [Errno 2] No such file or directory: '/home/sam/Downloads/SIT/KevalDoshi/ori_labels/2.cache'

[glenn-jocher]

@SamiurRahman1 this should be easy to do. Just point your coco128.yaml to your folder

val: inference/images/  # images

And run python test.py --save-json:

Namespace(augment=False, batch_size=32, conf_thres=0.001, data='data/coco128.yaml', device='', img_size=640, iou_thres=0.65, merge=False, save_json=False, save_txt=False, single_cls=False, task='val', verbose=False, weights='yolov5s.pt')
Using CPU

Fusing layers... Model Summary: 140 layers, 7.45958e+06 parameters, 7.45958e+06 gradients, 17.5 GFLOPS
Scanning images: 100%|██████████| 2/2 [00:00<00:00, 137.05it/s]
Scanning labels inference/labels.cache (0 found, 0 missing, 2 empty, 0 duplicate, for 2 images): 100%|██████████| 2/2 [00:00<00:00, 12557.80it/s]
WARNING: No labels found in inference/labels/. See https://github.com/ultralytics/yolov5/wiki/Train-Custom-Data
               Class      Images     Targets           P           R      [email protected]  [email protected]:.95: 100%|██████████| 1/1 [00:01<00:00,  1.14s/it]
                 all           2           0           0           0           0           0
Speed: 445.9/5.2/451.2 ms inference/NMS/total per 640x640 image at batch-size 32

Process finished with exit code 0

[SamiurRahman1]

@SamiurRahman1 this should be easy to do. Just point your coco128.yaml to your folder

val: inference/images/  # images

And run python test.py --save-json:

Namespace(augment=False, batch_size=32, conf_thres=0.001, data='data/coco128.yaml', device='', img_size=640, iou_thres=0.65, merge=False, save_json=False, save_txt=False, single_cls=False, task='val', verbose=False, weights='yolov5s.pt')
Using CPU

Fusing layers... Model Summary: 140 layers, 7.45958e+06 parameters, 7.45958e+06 gradients, 17.5 GFLOPS
Scanning images: 100%|██████████| 2/2 [00:00<00:00, 137.05it/s]
Scanning labels inference/labels.cache (0 found, 0 missing, 2 empty, 0 duplicate, for 2 images): 100%|██████████| 2/2 [00:00<00:00, 12557.80it/s]
WARNING: No labels found in inference/labels/. See https://github.com/ultralytics/yolov5/wiki/Train-Custom-Data
               Class      Images     Targets           P           R      [email protected]  [email protected]:.95: 100%|██████████| 1/1 [00:01<00:00,  1.14s/it]
                 all           2           0           0           0           0           0
Speed: 445.9/5.2/451.2 ms inference/NMS/total per 640x640 image at batch-size 32

Process finished with exit code 0

Thank you for the reply. This should be noted that the script only works when the images are in inference/images folder.
Hopefully one last question, i assume that it worked since i got the txt result files after running the following command:

python test.py --save-json --save-txt

So, where exactly is my .json file saved? I searched for any .json file in the folder, but didn't find anything. I got the following output on the terminal:

Namespace(augment=False, batch_size=32, conf_thres=0.001, data='data/coco128.yaml', device='', img_size=640, iou_thres=0.65, merge=False, save_json=True, save_txt=True, single_cls=False, task='val', verbose=False, weights='yolov5s.pt')
Using CPU

Fusing layers... Model Summary: 140 layers, 7.45958e+06 parameters, 7.45958e+06 gradients
Scanning labels /home/sam/Downloads/SIT/yolov5-master/inference/labels.cache (0 
WARNING: No labels found in /home/sam/Downloads/SIT/yolov5-master/inference/labels/. See https://github.com/ultralytics/yolov5/wiki/Train-Custom-Data
               Class      Images     Targets           P           R      [email protected]
                 all    2.68e+03           0           0           0           0           0
Speed: 141.7/4.2/145.9 ms inference/NMS/total per 640x640 image at batch-size 32

[DINOUS11]

@glenn-jocher Hi, after i output the result.json weird that it has many FP cases, is this file the final result with nms and confidence_score filter result already, btw thanks for this repo, nice work!!!

[glenn-jocher]

@DINOUS11 this is a pycocotools format json

[DINOUS11]

pycocotools format

@glenn-jocher Is that means it is given all prediction with nms and confident threshold filter?

[glenn-jocher]

yes

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.