Test model MAP: AssertionError: No labels found in \data\coco\labels\train2017 about cocopie-modelzoo HOT 9 CLOSED

Move your /coco folder to CoCoPIE-ModelZoo folder and then check it again.

Follow the instructions on README.md
Download Coco Dataset: (18 GB)
cd ../ && sh YOLObile/data/get_coco2014.sh

from cocopie-modelzoo.

I followed the same but it does not solve the issue.
I am getting the following error-

Traceback (most recent call last): File "test.py", line 263, in <module> opt.augment) File "test.py", line 59, in test names = load_classes(data['names']) # class names File "D:\Projects\AI360\CoCoPIE-ModelZoo-master\YOLObile\utils\utils.py", line 58, in load_classes with open(path, 'r') as f: FileNotFoundError: [Errno 2] No such file or directory: '../data/coco.names'

My project structure is now as below-

from cocopie-modelzoo.

oh so this might be a issue with windows I ignored. I do not have windows environment. I suggest checking the coco2014.data file in /data and change the path inside to windows-friendly.

from cocopie-modelzoo.

Thanks @nightsnack . Indeed it was path issue related to windows. I corrected it in .data file and now able to successfully run the test file.

from cocopie-modelzoo.

@danbader88 How were the results? is it the same? + Have you tried to write codes for demo etc?

from cocopie-modelzoo.

@MuhammadAsadJaved I tested the mAP for coco 2017 dataset in my NVIDIA GeForce GTX 1060 system and achieved the following-

But I have not tested the same on any mobile device yet.

from cocopie-modelzoo.

@dsbyprateekg Thank you for the results. Is there any suggestion to run the code for demo?

from cocopie-modelzoo.

@MuhammadAsadJaved You just need to follow author's README file and if you have windows machine then you need to adjust the path in .data file. It will run without any issue.

from cocopie-modelzoo.

@MuhammadAsadJaved You just need to follow author's README file and if you have windows machine then you need to adjust the path in .data file. It will run without any issue.

I am not talking about tests or comparisons. I am talking about checking the model on image/video inputs as a demo.

from cocopie-modelzoo.