Skip to content

Repo showing different ways/approaches for tracking objects

Notifications You must be signed in to change notification settings

mtc-20/Object_Tracking

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 

Repository files navigation

Object Tracking Algorithms

Compile and keep python scripts that can track objects.

[INFO_051219]: Haven't tested any of these codes on Raspberry Pi yet.

OpenCV Object Trackers

OpenCV has about 7 different object tracking algorithms, the accessibility to which depends on your version of OpenCV. Check out Adrian's guide to know more, since that's where I learnt about it. The code in this repo is also taken from the same guide.

Script

Notes:

  • The problem, atleast from my observations, is that they fail when the tracked object goes out of frame
  • So, these are more useful for follower applications, where the objective is to ensure the tracked object is always within the frame
  • The ROI GUI (again from OpenCV), although pretty neat, makes it somewhat inefficient for autonomous applications

GOTURN Object Tracker

GOTURN is a deep learning based object tracker, that is also available on OpenCV. For more info about the model as well as the trained model files, check out Satya Mallick's guide.

Model - Dropbox link

Script - GOTURN tracking

Notes:

  • The model fails/seems to break for objects or feed that the model was not trained for
  • If the object has gone out of frame for a moment, it seems to break
  • The major problem for me was that it's quite heavy in terms of resource usage; it made my PC laggy and pretty much crashed it (Ubuntu 18) the first (and only time) I used it. I havent tried running it on GPU yet