GOSPAMetric: Avoid unnecessary copies and inefficient masking #1059
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #1059 +/- ##
==========================================
+ Coverage 94.19% 94.23% +0.04%
==========================================
Files 207 207
Lines 13465 13487 +22
Branches 2740 2743 +3
==========================================
+ Hits 12683 12710 +27
+ Misses 530 524 -6
- Partials 252 253 +1
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. |
marvonlar
commented
Jul 16, 2024
| @@ -319,6 +350,8 @@ def compute_gospa_metric(self, measured_states, truth_states): | |||
| if num_truth_states == 0: | |||
| # When truth states are empty all measured states are false | |||
| opt_cost = -1.0 * num_measured_states * dummy_cost | |||
| if self.alpha == 2: | |||
| gospa_metric['false'] = opt_cost | |||
There was a problem hiding this comment.
This change was not part of my first patch, but came up as I added some tests to provide coverage of my original changes.
It seems reasonable that GOSPA should report tracks as false also in the case where there are 0 ground truth states, right?
I don't fully understand why we test for self.alpha == 2, but this is what is done everywhere else (see for instance line 369).
marvonlar
requested review from
orosoman-dstl and
mharris-dstl
and removed request for
a team
sdhiscocks
approved these changes
Jul 17, 2024
jswright-dstl
approved these changes
Jul 18, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
GOSPAMetric.compute_over_time()was taking a complete copy of all measurements and ground truth states for every timestep in the sequence. Additionally, it used multiple full linear masks for every iteration on the sorted timestamps. These steps cause the metric to haveThis PR performs the copying once and uses
itertools.groupby()to avoid the linear masking in each iteration. As a result,GOSPAMetricnow uses slightly less memory (we perform the same copy, but save the masking) and, in practice, runs linearly in the number of timesteps. The PR also includes a test which demonstrates 30-40x speedup on a 1k timestep long sequence. Offline I'm seeing 150x speedup (going from 20 min to 8 sec) on a real-world dataset with 50k timesteps.This patch is also likely to help in the case of #1032.