* continue to use paddleocr v3 text detection model for large
v5 was not finding text on multi-line plates at all in testing
* implement clustering of plate variants per event
should reduce OCR inconsistencies and improve plate recognition stability by using string similarity to cluster similar variants (10 per event id) and choosing the highest confidence representative as the final plate
* pass camera
* prune number of variants based on detect fps
* implement replacement rules for cleaning up and normalizing plates
* docs
* docs
* Cleanup onnx detector
* Fix
* Fix classification cropping
* Deprioritize openvino
* Send model type
* Use model type to decide if model can use full optimization
* Clenanup
* Cleanup
* Use OpenVINO directly to detect if devices are available
* Cleanup
* Update OpenVINO
* Cleanup
* Don't try to use OpenVINO when CPU is set as device
* Catch case where input tensor can't be pre-defined
* Cleanup
* Use re-usable inference request to reduce CPU usage
* Share tensor
* Don't count performance
* Create openvino runner class
* Break apart onnx runner
* Add specific note about inability to use CUDA graphs for some models
* Adjust rknn to use RKNNRunner
* Use optimized runner
* Add support for non-complex models for CudaExecutionProvider
* Use core mask for rknn
* Correctly handle cuda input
* Cleanup
* Sort imports
* update config for roles and add validator
* ensure admin and viewer are never overridden
* add class method to user to retrieve all allowed cameras
* enforce config roles in auth api endpoints
* add camera access api dependency functions
* protect review endpoints
* protect preview endpoints
* rename param name for better fastapi injection matching
* remove unneeded
* protect export endpoints
* protect event endpoints
* protect media endpoints
* update auth hook for allowed cameras
* update default app view
* ensure anonymous user always returns all cameras
* limit cameras in explore
* cameras is already a list
* limit cameras in review/history
* limit cameras in live view
* limit cameras in camera groups
* only show face library and classification in sidebar for admin
* remove check in delete reviews
since admin role is required, no need to check camera access. fixes failing test
* pass request with camera access for tests
* more async
* camera access tests
* fix proxy auth tests
* allowed cameras for review tests
* combine event tests and refactor for camera access
* fix post validation for roles
* don't limit roles in create user dialog
* fix triggers endpoints
no need to run require camera access dep since the required role is admin
* fix type
* create and edit role dialogs
* delete role dialog
* fix role change dialog
* update settings view for roles
* i18n changes
* minor spacing tweaks
* docs
* use badges and camera name label component
* clarify docs
* display all cameras badge for admin and viewer
* i18n fix
* use validator to prevent reserved and empty roles from being assigned
* split users and roles into separate tabs in settings
* tweak docs
* clarify docs
* change icon
* don't memoize roles
always recalculate on component render
* Update object_detectors.md for v16
* add configurability to IMG_SIZE for YOLOv9 export
* remove TensorRT detector as it's no longer supported in v16
* Revert removing NVIDIA TensorRT detector docs
Added documentation for NVidia TensorRT Detector, including model generation, configuration parameters, and example usage.
* Dumb copy/paste
* Enhance YOLOv9 export instructions in documentation
Updated YOLOv9 export command to include IMG_SIZE parameter and clarified model size options.
* model type does not need to match config model type
As long as a model is supported by a detector, it should be available in the list
* fix missing semicolon
the web linter was complaining
* Use asyncio lock when checking camera status
get_camera_status() can be called during normal autotracking movement and from routine camera_maintenance(). Some cameras cause one of the status calls to hang, which then subsequently hangs autotracking. A lock serializes access and prevents the hang.
* use while loop in camera_maintenance for status check
some cameras seem to take a little bit to update their status, don't assume the first call shows the motor has stopped
