GithubHelp home page GithubHelp logo

Tracking Parameters about multitarget-tracker HOT 7 CLOSED

isra60 avatar isra60 commented on August 16, 2024
Tracking Parameters

from multitarget-tracker.

Comments (7)

Nuzhny007 avatar Nuzhny007 commented on August 16, 2024 1

Hi!
For this example approach with Background subtraction is not a good choice.
But if BGFG segmentation is used then for better performance it need to disable KCF and other trackers.
I think than approach with deep_sort and your with YOLO ( #13 (comment) ) may be better. I want to implement it.

I'm using BGFG segmentation for big background areas and small foreground objects.

from multitarget-tracker.

isra60 avatar isra60 commented on August 16, 2024 1

Maybe using SSD is better than YOLO. Based on https://www.pyimagesearch.com/2017/09/11/object-detection-with-deep-learning-and-opencv/.
I don't exactly what deep_sort tracking does. I guess that detects rectangle based on YOLO detector but I donw know how it follow the track

from multitarget-tracker.

Nuzhny007 avatar Nuzhny007 commented on August 16, 2024

Hi!

dt - delta time. It's parameter for Kalman filter, model time interval

dist_thresh - threshold for matching tracked targets on video and detected regions on the current frame. If used DistCenters then dist_thresh - Euclidean distance between centers (in pixels). If used DistJaccard then dist_thresh - Jaccard disnance (from 0 to 1).

maximum_allowed_skipped_frames - If the target was not found at the specified number of frames, it will be deleted. It can use fps to set this parameter in seconds: maximum_allowed_skipped_frames == 2 * fps. Tracking the target 2 second without detections.

max_trace_length - Each target contains trajectory with this length.

from multitarget-tracker.

Nuzhny007 avatar Nuzhny007 commented on August 16, 2024

accelNoiseMag - this parameter for Kalman filter, it's noise.

With work of the dt and accelNoiseMag you can experiment in MouseTracking example (-e=0).

from multitarget-tracker.

isra60 avatar isra60 commented on August 16, 2024

Thanks. Any hints of how can I handle occlusion?? I know kalman filter is the way to it but I don't get good results. when the target pass the occlusion it gets a new id..

from multitarget-tracker.

Nuzhny007 avatar Nuzhny007 commented on August 16, 2024

Mmm...
Kalman can track a target when it's fully occluded. KCF can help if target partial occluded.
May be you can show video with this problem?

from multitarget-tracker.

isra60 avatar isra60 commented on August 16, 2024

For example. Maybe its a bit challenging this video.

https://drive.google.com/file/d/1lS2HMJKFxryBSwoCDX--u-wujiv9rxeD/view?usp=sharing

Which parameters should we change in order to get better performance of the tracking??

from multitarget-tracker.

Related Issues (20)

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.