YOLOv8 making false positives detections

Question

I have trained 215 pics which are contains chess pieces with using yolov8m.pt as pre-trained model and epoch value was 75.

!yolo task=detect mode=train model=yolov8m.pt data={dataset.location}/data.yaml epochs=75 imgsz=800 plots=True

Then when I tried that model it detected stuff which is not in train set. How can I solve that problem.

score 0 · Answer 1 · answered Jul 20 '23 at 07:42

0

The manner in which you can solve this problem is to add images to your dataset such as the ones above/similar, in which there are no labels of interest.

If you only trained on images and labels from the chess training pieces, it's likely that the model would yield false positives such as the ones above.

For both YoloV5 and YoloV8, it is recommended that you also include background images(images with no labels) which could amount from 1% to 10% of your entire dataset.

answered Jul 20 '23 at 07:42

Timbus Calin

13,809
5
41
59

How can I add background in a dataset. Which name should be background label? – gurselturkeri Jul 20 '23 at 07:54
There should be no written lines in the label files (empty text file). You should have the image such as image_name.jpg and the image_name.txt, where the image_name.txt is an empty file. – Timbus Calin Jul 20 '23 at 08:20
Did you understand @gurselturkeri? – Timbus Calin Jul 20 '23 at 12:47

YOLOv8 making false positives detections

1 Answers1