Appearance
What's New in 2.9
Welcome to FORXAI Video Vision 2.9!
Summary
This minor release brings support for reactively changing machine learning model used based on inputs from user or automation allowing detection of multiple product types without disruptions to production lines.
Additionally ability to trigger Basler camera by integrating it with other operational technologies such as laser gates allows processing of selected camera frames instead of continuous detection.
See more details about hardware triggers in dedicated section. Read into ability to dynamically change models here and see more updates on developed logic to support this function in section about nodes.
See information about hotfix updates:
Major Features
Major features introduced in this release are described below. This release includes other changes related to fine-tuning FVV capabilities, bugfixes, UX and other improvements which are omitted in this article.
Basler camera hardware trigger
Added support for a hardware trigger, an external signal that controls when the camera captures an image. This ensures precise synchronization in applications such as defect detection, conveyor belt tracking, and high-speed monitoring. By using a hardware trigger, image capture can be aligned with external events, reducing latency and improving accuracy in time-sensitive workflows
Dynamic model switching
Added support for Dynamic model switching, enabling seamless transitions between AI models at runtime without requiring a system restart. This enhances flexibility in environments where different models are needed for different tasks, reducing downtime and manual intervention.
For example, a production line handling multiple product types can automatically switch between defect detection models based on the item being processed.
New nodes
Dynamic AI Model
The Dynamic AI Model node implements the dynamic model switching capability within the workflow. It allows users to configure multiple machine-learning models and swap between them as needed without restarting the deployment.
Unlike the standard AI Model node, which runs a single preloaded model, the Dynamic AI Model node can dynamically load and unload models, making it ideal for workflows that require adaptability—such as detecting defects across different product categories or optimizing resource use when GPU memory is limited. Learn more.
Overlay Heatmap
The Overlay Heatmap node overlays a heatmap onto an image, allowing you to visualize data directly within the image context. The node includes options to adjust the heatmap's opacity and select from several predefined color gradients. Learn more.
Fixed issues
- Significantly improved the load times of Reporting entries by optimizing data storage and retrieval.
Compatibility
The Video Vision platform was tested on Ubuntu Server 22.04.5 LTS. No breaking changes were identified.
This version powers minor release of Forxai Mirror v2.5.0.
Changes in hotfix 2.9.1
Fixed compatibility issue between Dynamic AI Model Node and Anomaly Detection models.
Changes in hotfix 2.9.2
Model outputs were not correctly propagated when prediction probability was 0