New vectorized distance functions. #211
Merged
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
facundo-lezama
force-pushed
the
vectorized-distance-functions
branch
9 times, most recently
from
2549262 to
0815f94
Compare
javiber
requested changes
Nov 9, 2022
7 tasks
|
Why do we keep both What use case does the unoptimized version have? |
@dekked the unoptimized version has some validations that might be useful. This PR doesn't address the modification of the IoU distance function as a vectorized distance, that's why we don't modify it here. In #213 we are changing the optimized IoU implementation, and we will revisit these two definitions. |
facundo-lezama
force-pushed
the
vectorized-distance-functions
branch
from
74700c0 to
a98a86a
Compare
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Outdated suggestions cannot be applied.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.
In this PR we add support for vectorized
distance_functionsto speed up Norfair. The new distances can be one ofScalarDistance,VectorizedDistance, andScipyDistance. We moved the distance functions and the distances calculations (previouslyTracker._get_distances) to the newly mentioned classes. With this new interface, we can define a variety of distances; from simple scalar distances such as the ones that we have been using until now, to custom vectorized distances that take a pair ofnp.ndarrayas input, and even default vectorized distances fromScipy.We were careful not to break the API. The users can still define custom distances and now they can also leverage predefined vectorized distances that speed up the tracking.
Changes include:
distances.py.Tracker._get_distancesand moved it to the distance classes.Tracker.__init__to create instances of the new distance classes.Pending tasks:
[ x ] Change demos to use vectorized distances when possible
[ x ] Check ReID still works
[ x ] Add tests
[ x ] Review and update any deprecated documentation
[ x ] Write PR description