Robovision 5.8: Algorithm Upgrades and Improved Vision Capabilities

YOLOv8+: Compare a Proven Algorithm to the Latest Innovation

Robovision’s 5.8 implementation of the YOLOv8 (You Only Look Once) algorithm streamlines both instance segmentation and object detection processes, which previously used SOLOv2 and YOLOv5. In comparison, YOLOv8 makes training time 10x faster and inference time 5-10x faster, while giving more precise outputs.

Much discussion surrounds the YOLO releases in the data science community. Some practitioners express skepticism about the frequent release of new YOLO versions, noting that each iteration offers only marginal improvements. Others appreciate the continuous enhancements, particularly in speed and efficiency, acknowledging that while accuracy gains may be incremental, the reduced training times and resource requirements are still valuable. 

With the v5.8 release, Robovision makes the tools available for data scientists to experience the differences between algorithms themselves. 

YOLOv8 is the primary engine driving instance segmentation and object detection with Robovision 5.8, but also implemented as a sort of extension of YOLOv8 is the ability to compare it directly with YOLOv11. 

The development of models using both v8 and v11 algorithms is made possible—allowing for direct comparison of the results using both. Depending on the specific use case, one might outperform the other. Whatever the result, the freedom to implement the algorithm that delivers the best results is available.

At the very least, this function provides data scientists with first-hand experience, enabling them to actively participate in current and relevant discussions within the community.

Seeing is Believing: Enhanced Visualization and Data Capture

Robovision 5.8 introduces significant improvements in data capture and new capabilities for leveraging that data to deliver outcomes. 

Live Camera Feed

Latency is always an issue when dealing with real-time processing of large amounts of data. Previously, images and data captured through the camera were inaccessible. The only visuals available on the human side were data representations in charts and graphs. 

For the 5.8 release, WebRTC technology and peer-to-peer (P2P) network connections allow direct data transmission between devices without facing major delays due to latency. In other words, real-time data capture through live camera streaming is now available, so operators can see the images and data classification happening on the line in real time. 

Company-specific GUIs can also be created, allowing the development of bespoke dashboards tailored to specific needs. This allows for more immediate and informed decision-making.

Improved Management of AI Model Output

The new 5.8 release’s UI can be used to integrate AI model output with other systems. Link a camera to the Robovision software, for example, and integrate it into inference pipelines to start capturing valuable data.

From there, send inference results to a webhook—or “digital trigger”—which allows Robovision’s computer vision AI software to automatically communicate with another software application. Once the AI model identifies an object in an image, it can send a notification to another system that can perform various actions, including: 

• Saving the image so it can be analyzed more thoroughly later. 
• Sending an alert notifying a human operator in case of (for example) a security breach or a quality issue.
• Triggering another process or automated task, such as activating a machine or sending a message using web apps or cloud-based systems.

Setting up a webhook triggers an automated response to the model’s predictions, making more efficient processes. 

Along with more control over captured data, a new metadata feature automatically tracks information like camera type, serial number, resolution, timestamp, and name for each recorded sample. Two fields are left blank and can be used to fill in custom metadata—logs can be downloaded for further analysis down the line.

Big Upcoming Plans

Some of the innovations that are still in development are so extensive they require more than a single quarter to fully perfect and integrate. To ensure every new feature meets the highest standards, rigorous testing and development continues. While Robovision 5.8 brings some majorly transformative features, its true purpose is to serve as a foundation for even bigger plans ahead.