ARL Research Project - Deev Patel

Goal

Create application to identify and track unclassified objects

Building with CMake

mkdir build
cd build
cmake ..
make

Executables

• Basics from openCV's abilities
  • ./old_regions
    • For viewing the different region proposal methods
  • ./old_saliency
    • For viewing saliency model outputs
  • ./old_contours
    • For viewing OpenCV contour seqmentation algorithm
  • ./old_filteredRegions
    • For testing out region filtration methodology
• YOLOv3
  • ./yolo_visualization
    • For visualizing the output of YOLOv3 at all 3 levels
  • ./yolo
    • C++ interface to YOLOv3
• Segementation
  • ./cv_segmentation
    • For running openCV selective segmentation
  • ./segmentation
    • For running modified segmentation algorithm
  • ./segmentation_old
    • For running old version of modified - merging down without regard to level of origin
• Video Testing
  • ./webcam
    • For running YOLOv3 & openCV selective segmentation on webcam
  • ./fly_capture_camera
    • For running YOLOv3 & modified segmentation on flycapture camera
  • ./fly_capture_camera_detailed
    • For running YOLOv3 & modified segmentation in debug mode (still displaying old domains) on flycapture camera
• ./test_area
  • For testing experimental stuff - selective segmentation & saliency

YOLOv3 Interface

The interface to YOLOv3 is linked to the darknet library. The library is created (in the build directory) from the darknet source files ("/src/yoloInterface/darknet") and the darknet makefile. So, to update the darknet library after changing its source files, rebuild the entire project. By default, darknet is built with OpenMP, Cuda, and CUDNN.