Comments (3)
Let me clarify. In py-motmetrics the threshold you mention to keep associations alive is not an explicit parameter. Instead, it is part of the given distance matrix. All elements that cannot pair are signalled with NAN.
What py-motmetrics does, it allows you to compute distances between hypothesis of your tracker and ground truths of the current frame. It then prevents potential pairings (of established or new tracks) that exceed a given threshold via indication with NAN in the input distance matrix.
Why not a making a parameter? Depending on the application at hand you might want to apply different thresholds to different objects based on some higher level logic (think for example: your tracker outputs a confidence for each hypothesis). This is unattainable with a global parameter, but easily achieved by appling different NAN rules to different objects/hypothesis.
So, [0] 2.1.1 states
First of all, the correspondence between an object oi and
a hypothesis hj should not be made if their distance disti,j
exceeds a certain threshold T.
This statement does not relate to tracks already established.
In [2] it seems like a greedy assignment algorithm is used on the principle of first-come-first-serve. In contrast py-motmetrics assigns the remaining possible pairs to matches, misses and switches using a per-frame optimal algorithm based on minimum costs (LAP). Where remaining refers to all potential pairs, after already established ones have been continued.
Re-establish:
https://github.com/cheind/py-motmetrics/blob/master/motmetrics/mot.py#L171
Assign remainings:
https://github.com/cheind/py-motmetrics/blob/master/motmetrics/mot.py#L190
Does this clear up things for you?
from py-motmetrics.
Mmm I see, so it is something which has to be done before we fill in the distance matrix. Also, I realized I completely misunderstood 2.1.1 in the paper, sorry about that :D
from py-motmetrics.
Correct. You are welcome, I had to look it up too! ;)
from py-motmetrics.
Related Issues (20)
- MOTP with euclidean distance metric gives mean squared distance. HOT 1
- how to solve this problem? HOT 1
- eval_motchallenge.py: error: the following arguments are required: groundtruths, tests
- Total number of objects calculation HOT 3
- Do we need to keep original detection order? HOT 5
- About OSPA/GOSPA HOT 2
- Wrong results HOT 1
- Everything gets 0 / NaN if confidence is -1 in motChallenge format HOT 1
- AttributeError for `np.bool` in Numpy `1.24.0` HOT 5
- How to calculate IDF1 metric for multi-camera multi-target(MCMT,MTMC) tracking? HOT 1
- MOTP Metric
- Precision-Recall Curve
- Why idp and idr samein result?
- Custom dataset example in `README` gives worse results than `motmetrics.apps.eval_motchallenge`
- A BUG caused by pandas.Index.map
- can idp and idr be more than 1? HOT 1
- Person exits and re-enters
- Questions about values HOT 2
- mostly_tracked(MT) HOT 2
- Error with Default Option "ana" When Computing Metrics
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
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.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from py-motmetrics.