• Helm Chart Version: 2.0.3
• Alembic Version: 2906615bb9c3

Features

• Automatically recommend VM models based on detection scenarios: A function has been added to automatically recommend a Vision Model optimized for detection scenarios. This reduces the user burden on model selection and ensures stable detection performance for each scenario.
• Polygon area setting function support: Improved to allow free polygon-based area setting beyond the limitations of the existing rectangular area setting. More precise detection area configuration is possible even in complex spatial structures.
• n8n Webhook Channel Expansion: The channel that can be linked to external systems such as PAT using n8n webhook has been expanded. This allows more flexibility in configuring notification linkage scenarios.
• Added screen brightness adjustment function: A screen brightness adjustment function has been added to maintain stable detection performance even in various lighting environments. The slider allows you to set it in the range -100 ~ 100, with negative numbers making it darker and positive numbers making it brighter.
• Activation of user permission feedback function: The detection notification feedback input function, which was only possible with existing administrator (admin/manager) permissions, has been extended to user permissions.
• Added feedback input cancellation function: A function has been added to allow the user to cancel input after inputting feedback for a detection notification.
• Select all favorites function provided: The select all function has been added so that you can register the filtered camera list as favorites at once.
• Provision of feedback filter for viewing detection notification history: When viewing detection notification history, a function has been added to select the feedback type as a filter.

Performance Enhancements

• Improved object detection sensitivity: We minimized setup deviation by normalizing the sensitivity standard value for each model. Improvements have been made to ensure that existing user settings are maintained even when changing models.
• Webhook setup UX improvement: The input order has been changed to enter the webhook URL first and then enter the name, as in the camera registration flow.
• Improved Camera Settings RTSP URL UI: In the Camera Settings screen, the component where the RTSP URL appeared to be editable was changed to non-editable to minimize user confusion. However, the URL copy function is maintained.
• Improved camera detail screen structure: The settings areas distributed for each camera have been reorganized.
  • The top area of the streaming has fixed setup menus for each camera (camera information, detection scenario, detection target & area, channel notification).
  • Screen brightness adjustment and deletion settings provided through the three dots (⋮) button
  • The information area at the bottom of the streaming provides detection scenario, detection target, model information, and inference cycle information in a unified form.

UI / Branding

• Mellerikat logo change: The existing mellerikat EVA logo has been changed to the EVA logo.

Bug Fixes

• Webhook Validation issue fixed: Fixed an issue where the validation status was reset when entering a name after webhook URL validation.