YOLOv5 training data prompts No labels found, with_suffix is ​​used, WARNING: Ignoring corrupted image and/or label appears during yolov5 training

YOLOv5 training data prompts No labels found [Pro-test yolo to load the label file only needs to modify the img2label_paths function changes to load the label file]

Looking carefully at the file datasets.py for data loading and processing, I found that there is a sentence that will find the corresponding labels folder according to the location of the images folder in step 2:

The place where YOLOv5 loads the label is in this place in datasets.py, we just modify the path to load the label to place the label for our own.

In this img2label_paths function, our modifications are as follows: [Because we put the label and img in the same folder, we can just modify the suffix name directly]

with_suffix(suffix) replace the extension, return the new path, the extension will remain unchanged

You can also use with_suffix(suffix) to modify the suffix, but this method needs to import the Path('/hidog/text.tar.gz') class,

This is used in YOLOv5, used when checking the cache file

• [The cache will continue until the check data file is executed, which contains nc, nf, etc., which can display whether the data is found, lost, damaged, etc. on the console. For example, when the input image is a raw image that cannot be processed by yolo, it will be nc+=1 accumulated, and finally printed in the console]

Use of .replace() function in YOLOv5

yolov5 is used like this

WARNING: Ignoring corrupted image and/or label during yolov5 training

When using the yolov5 training data set, the following alarm appears

WARNING: Ignoring corrupted image and/or label

We found the place where the warning was found in the debug order

The reason is that the image I input is in the .raw format that YOLOv5 cannot process. Here we choose to convert .raw to .png or .tif image and then upload it.

Reference: Python standard library pathlib

Reference: WARNING: Ignoring corrupted image and/or label appears during yolov5 training
YOLOv5 training data prompt No labels found