studyproject_machine_learning

RGB Group

next steps

• distribute tasks
• shift bounding boxes to RGB stream
• implement preprocessing ideas
• parallel: train YOLO and compare results
  • train on raw RGB stream -> implement pipeline
  • after finished pipeline: train on preprocessed images

ideas preprocessing

• picture subtraction
• background subtraction
  • KNN-bs
  • MOG2
• convert RGB in greyscale or HSV
• delete different bands -> try and error (lower priority)
• neighborhood operations
  • edge detection
• optical flow

folder structure

• 16mm_dvs-30_basler-70_combined_veryslow
• insects
  • train
    • images
    • images_rgb
    • labels
    • labels_rgb
  • val
• scripts
  • preprocessing
  • computer vision operations
  • yolo scripts ?
• videos
• yolo7