• Helm Chart Version: 1.6.1
• Alembic Version: 3599f5503a55

Features

• Introduction of licensing function: A new function has been added to control the use and activation of EVA solution functions based on license. To use EVA, license key issuance and registration procedures are required.
• Automatic blurring of human faces: Automatically blurs human faces in detected images, and the blurred images are applied to the detection DB, chat window notifications, and webhook transmitted images.
• UI & UX improvements
  • Added function tooltips: Tooltips have been provided for the Edit, Webhook, and Delete buttons on the camera detail page to improve understanding of the function.
  • Improved detection area settings: The default label has been set to ‘Area’ and the streaming screen has been improved to be immediately visible even without a label.
  • Change of detection notification information: Modified the detection notification details screen to display eva's response instead of image description.
  • Improved object detection sensitivity: We changed the term ‘detection sensitivity’ to ‘object detection sensitivity’ to clarify the meaning, added a guide, changed the adjustment range to 0.1 units, and improved automatic selection when adding a target.
  • Additional step-by-step saving function: A Save button has been added to each step on the camera detailed settings page, allowing step-by-step saving of settings.
  • 'AI Inference Interval' name overhaul: Changed 'Detection Interval' to 'AI Inference Interval' to clarify the meaning.
  • Improved feedback author information: The ID of the last feedback inputter has been changed to display the full ID instead of an abbreviation.
  • Added detection scenario guide: We have added guides to the common scenario and custom scenario setup areas to make the configuration flow easier to understand.

Performance Enhancements

• Booting speed improvement: To solve the DB access concentration problem that occurred when multiple cameras (more than 100) are running simultaneously, the coalesce technique that bundles and processes events was applied to reduce the event loop delay in the booting process.
• Optimization of DB usage structure: The structure was changed to a single DB approach when starting the camera pipeline, improving it to a form suitable for service separation and pipeline microservice. As a result, DB load is reduced and overall processing performance is improved.

Bug Fixes

• Correction of language setting transfer error: We have corrected the issue where the app's language settings were not accurately transferred to the Agent during the detection scenario generation process and improved the prompt processing for each language to operate normally.
• Fixed an issue where user input values were overwritten: Fixed an issue where automatically updated data was overwritten on the user input values when switching pages or returning to the screen.
• Fixed camera list page movement error: We have fixed the issue of returning to page 1 of the list when viewing more than 2 pages in the camera list → entering the camera details screen → pressing back, so that the position of the page the user was viewing is maintained properly.
• Fixed issue of excessive debugging log output: Fixed an issue where the debugging log output when error 422 occurred during a Vision ML API request included an image frame value, resulting in an excessive increase in log capacity.
• Improvement of broken UI when zooming in: Fixed a UI issue where the table appeared broken when increasing the screen magnification on the camera details page.