Skip to content
No description, website, or topics provided.
C++
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
README.md
detect_and_track.cpp
tracker.cpp
tracker.h
yolo_v2_class.hpp

README.md

rt_boat_tracker

Real Time Boat Tracker

Description

The tool detects and tracks special boats from live stream or video file.

Detection and tracking techniques

The project is based on a YOLO object detection model. A YOLO model has been trained which detects special boats.

The tracking algorithm uses the trained model to detect boats on each video frame. Then it compares detections with the detections founded in previous frame and compares the relative locations of the objects considering objects as the same if they are near to each other.

Dependencies

  • YOLOv3
  • OpenCV

How to build and run

  1. Clone Windows and Linux version of Darknet YOLOv3 and build with GPU support.
  2. Locate your boat detection model (YOLO .weights file) in the build\darknet\x64\backup directory and the YOLO .cfg file in build\darknet directory.
  3. Clone rt_boat_tracker inside the build\darknet directory, open build\darknet\yolo_console_dll.sln in MSVS2015 and replace the yolo_v2_class.hpp and yolo_console_dll.cpp files with the rt_boat_tracker checkouted files.
  4. Build the project.
  5. Run the .exe file build\darknet\x64\yolo_console_dll.exe and enter video file name.
You can’t perform that action at this time.