Mistaken path change from "images" to "labels"

[tetelevm]

Hello! I create dir /images/train/ on my GoogleDrive and put the train images into /images/train/images/ and labels into /images/train/labels/. I expect the script to look for labels in /images/train/labels, but it looks at the /labels/to_train/images/ (replaces the first word "images" instead of the last word). Isn't this a bug?

[glenn-jocher]

@tetelevm yes I see the same result in master. Code to reproduce:

import os
from utils.datasets import img2label_paths

img2label_paths(['dataset/images/images/img1.jpg'])

# output ['dataset/labels/images/img1.txt']

[glenn-jocher]

Possible fix:

import os

def img2label_paths(img_paths):
    # Define label paths as a function of image paths
    sa, sb = os.sep + 'images' + os.sep, os.sep + 'labels' + os.sep  # /images/, /labels/ substrings
    return ['txt'.join(x.replace(sa, sb, 1).rsplit(x.split('.')[-1], 1)) for x in img_paths]


def img2label_paths_fix(img_paths):
    # Define label paths as a function of image paths
    sa, sb = os.sep + 'images' + os.sep, os.sep + 'labels' + os.sep  # /images/, /labels/ substrings
    return ['txt'.join(sb.join(x.rsplit(sa, 1)).rsplit(x.split('.')[-1], 1)) for x in img_paths]


def img2label_paths_fix2(img_paths):
    # Define label paths as a function of image paths
    sa, sb = os.sep + 'images' + os.sep, os.sep + 'labels' + os.sep  # /images/, /labels/ substrings
    return [sb.join(x.rsplit(sa, 1)).rsplit('.', 1)[0] + '.txt' for x in img_paths]

images = ['dataset/images/images/img1.jpg'] * 100

print(img2label_paths(images[0:1]))
print(img2label_paths_fix(images[0:1]))
print(img2label_paths_fix2(images[0:1]))

%timeit img2label_paths(images)
%timeit img2label_paths_fix(images)
%timeit img2label_paths_fix2(images)

## OUTPUT
['dataset/labels/images/img1.txt']
['dataset/images/labels/img1.txt']
['dataset/images/labels/img1.txt']

55.2 µs ± 519 ns per loop (mean ± std. dev. of 7 runs, 10000 loops each)
70.7 µs ± 2.09 µs per loop (mean ± std. dev. of 7 runs, 10000 loops each)
51.8 µs ± 336 ns per loop (mean ± std. dev. of 7 runs, 10000 loops each)

It looks like we have a winner.

[glenn-jocher]

@tetelevm good news 😃! Your original issue may now be fixed ✅ in PR #3720. To receive this update:

• Git – git pull from within your yolov5/ directory or git clone https://github.com/ultralytics/yolov5 again
• PyTorch Hub – Force-reload with model = torch.hub.load('ultralytics/yolov5', 'yolov5s', force_reload=True)
• Notebooks – View updated notebooks
• Docker – sudo docker pull ultralytics/yolov5:latest to update your image

Thank you for spotting this issue and informing us of the problem. Please let us know if this update resolves the issue for you, and feel free to inform us of any other issues you discover or feature requests that come to mind. Happy trainings with YOLOv5 🚀!