search

AlexeyAB / darknet

YOLOv4 / Scaled-YOLOv4 / YOLO - Neural Networks for Object Detection (Windows and Linux version of Darknet )
http://pjreddie.com/darknet/
Other
21.66k stars 7.96k forks source link

I don't want to apply augmentation with my dataset. #8236

Closed Wooks-git closed 2 years ago

Wooks-git commented 2 years ago

I trained yolov4-csp-custom (width,height = 256, classes=3) But when i using $./darknet detector train .data .cfg .weights -dont_show -show_imgs command, i got strange annotation. When i check my annotation files (e.g img1.txt , img2.txt ... files) they were not problematic. So i want to check my input data without augmentation.

  1. Does anyone know how not to apply augmentation?
  2. The annotation files are normal, but does anyone know why the input data is weird?
thuangb commented 2 years ago

Just use labeling tools such as yolo_mark, labelimg, ... to check you annotation

stephanecharette commented 2 years ago

FAQ: https://www.ccoderun.ca/programming/darknet_faq/#turn_off_data_augmentation

Wooks-git commented 2 years ago

Sorry, there was a mistake in the process of making the label file.