SequenceTimeOfNextCapture was reporting time with respect to two
different frames depending on context: either it would report absolute
time since the start of the sequence, if the first capture had not been
completed, or it would report the amount of time left until the next
capture, relative to the current time.
Removed logic which reported relative time, which seems like it was not
part of any code path, or was being corrected for elsewhere in
SimulationState's update loop. Removed some of the error correction in
the update loop and added stricter expectations for the management of
the timing code.
Added unit tests to ensure function reports time we'd expect regardless
of context.
Created follow-up task AISV-845 for making the timing code less fragile.
* changed the namespace of randomization tools to experimental
* moved randomization UI code to experimental namespace
* updated docs to emphasize that the randomization tools are experimental
* updated changelog
* Merge branch 'dr-test3' of https://github.com/Unity-Technologies/com.unity.perception into dr-test3
Add some tooltips and UI labels to perception camera, labelers, randomization scenarios, and parameter configurations
* making requested change
variable name case change
* added ///<inheritdoc/> to description overrides
* two more xmldoc additions to fix warnings
* Fixing labeling when renderers have per-material MaterialPropertyBlocks. Also supporting refreshing labeling manually after making changes to renderers or the Labeling component.
* updating changelog
* Lens Distortion now added as a pass for both semantic and instance segmentation by way of a LensDistortion shader. This can be enabled either through enabling post-processing in URP or HDRP respectively or by utilizing the override although this approach is internal only currently but allows for the manipulation of the shader without having to have access to the global volume which in some test scenarios is not available.
* Upgrading projects and dependencies
* Adding additional editor versions for testing.
* Adding test ensuring datasets are generated on exiting play mode.
* Reducing test set for per-commit to a single editor. Removing use of IncludeTestAssemblies from BuildPlayer* tests to work around mysterious issue with shader compilation.
* Fix yaml
* Adding "All Tests" which triggers nightly
* Switching to newer 2019.4
* Fixing instance and semantic segmentation shaders in 2020.1. Bumping package version. Removing publish tests from all tests and renaming
* Promotion test rename
* Update to latest capture and core. Switch off 2020.2 testing because it is not currently supported. Switch CI back to 2019.4.6f1 due to linux editor import hang.
* Updating changelog
* Do not report bounding boxes for objects that do not match the label config.
* Add test.
* Updating changelog
* Attempting to fix test
* Removing editor platforms for a test that gives unstable output in editors
* Do not generate any dataset if the PerceptionCamera is never enabled and no other dataset generation APIs are called.
* Fixing style and performance issues
* Multiple object editing
* Remove indents from inspector
* Fix to commit changes on a prefab labeling edit
* Responsive label editing
Fix so that add label does not lock up
* Labeling editor fixes
* Remove placeholder text
Hopefully only temporary, but it seems to be almost impossible to get placeholder text to work in current gui implementation
* Update visual UI to use IMGUI system
Rework of visualization to use IMGUI system, allows for UI rendering to occur after all labelers execute.
* Fix visualization off check
* Fix for segmentation images responding to screen size changes
* Adjustments for PR requests. Fixed tests in HDRP.
* Move get uniqueLabelerId to public to fix editor tests
* Changed HUD to just use the labeler as the key
* Display no labeler enabled message
If perception is ran with show visualizations on and no labelers active, a message box will display saying that no labelers are currently active
* Segmentation Alpha Bug Fix
Capped the max opacity allowed in the shader to an extremely high value (0.995) to stop a weird issue caused by alpha being set to 1.0
* Segmentation Alpha Bug Fix
Capped the max opacity allowed in the shader to an extremely high value (0.995) to stop a weird issue caused by alpha being set to 1.0. Incorporated the suggestion on how to clean up shader code
GitHub
Steven Leal
mohsen.k
Devin Miller
christopher.kerr
Steven Borkman