Lazy Masks / Detections2D #979
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
added 6 commits
January 21, 2022 11:56
…ate masks for all objects
…htX32bit int with zero copying
|
This pull request introduces 1 alert when merging 4da7e68 into 1d35fba - view on LGTM.com new alerts:
|
|
This pull request introduces 1 alert when merging c68795e into f9fb7fa - view on LGTM.com new alerts:
|
|
This pull request introduces 1 alert when merging 84d4f16 into f9fb7fa - view on LGTM.com new alerts:
|
|
This pull request introduces 1 alert when merging 1dacd77 into ea60106 - view on LGTM.com new alerts:
|
|
This pull request introduces 1 alert when merging f518057 into dc0f9ec - view on LGTM.com new alerts:
|
|
This pull request introduces 1 alert when merging 27ac088 into dc0f9ec - view on LGTM.com new alerts:
|
Lucaweihs
reviewed
Apr 19, 2022
added 7 commits
April 20, 2022 13:02
… all the instancedetections and then an iter()
|
This pull request introduces 2 alerts when merging a262b25 into 86e58cb - view on LGTM.com new alerts:
|
Lucaweihs
approved these changes
Apr 21, 2022
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Outdated suggestions cannot be applied.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.
Currently, when using instance segmentation (
ai2thor.controller.Controller(renderInstanceSegmentation=True)) instance masks, class masks and 2D detections are generated for every object in the image. This process is quite costly and end-users don't often use every single bounding-box or mask that is generated. This PR introduces lazily generated masks and 2D detections. In the absolute best case, users will get a 4x speed improvement (this means that no masks or 2D detections are generated, but the instance segmentation frame is sent from Unity to Python). In the worst case (we generate all the bounding boxes and masks) users should see a 2x improvement over the current implementation. This speed improvement is largely due to the elimination of the use ofnp.uniqueandnp.tilein favor of a zero-copy based approach. The relevant properties onEventare:instance_masks,class_masks,instance_detections_2D, andclass_detections_2D. All these properties now point to a class that implements theMappinginterface, but the behavior should be identical to the current scheme (i.e. if a key doesn't exist in the lazy-map a KeyError will get thrown). My plan is to merge this in post-ECCV.