EverFocus will present live demos powered by NVIDIA technologies at COMPUTEX 2026, highlighting edge AI vision, Sarea Aegis analytics, and smart factory inspection applications.

The live demo area will show how NVIDIA accelerated edge AI technologies can support real time object detection, robotic control, multi stream video analytics, semantic video search, and industrial inspection workflows.

Live Demo 1 | Robotic Arm Sorting & Stacking

Powered by NVIDIA Jetson Thor, the system uses real time AI vision to detect objects, estimate their 3D position, and guide the robotic arm through picking, sorting, and stacking tasks. The demo highlights how edge AI can support automated factory operations with faster visual recognition and precise robotic control.

 

Figure 1. Robotic arm sorting & stacking

Demo Capabilities

• Real time object detection and picking
• 3D positioning for robotic arm control
• AI vision based object recognition
• Robotic sorting and stacking workflow
• Automated factory vision 

NVIDIA Technologies Applied

Powered by NVIDIA Jetson Thor, this demo uses NVIDIA CUDA and NVIDIA TensorRT for accelerated computing and AI inference, supporting real time object detection and 3D positioning. NVIDIA Isaac ROS connects the AI vision pipeline with ROS 2 based robotic arm control, so detected object positions can be converted into picking, sorting, and stacking actions.

Figure 2. EverFocus EAR-100T powered by NVIDIA Jetson Thor

For advanced motion planning, NVIDIA Isaac ROS cuMotion adds GPU accelerated path planning, helping the robotic arm plan movements with better speed, accuracy, and control in factory automation workflows.

Live Demo 2 | Smart Factory Inspection using Hexapod Robot

The hexapod robot patrols the inspection area and streams live video to the Sarea Aegis platform. The Sarea Aegis then performs real time video analytics, including object detection, heatmap generation, event analytics, semantic search, and video summarization.

 

This demo shows how a mobile robot can become part of a smart factory inspection system, helping operators monitor restricted areas, detect events, review video data, and search visual information more efficiently.

 

Demo Capabilities

• Hexapod robot factory patrol
• Smart factory patrol and inspection workflow
• Multi stream video analytics
• Real time object detection
• Heatmap generation
• Semantic video search and video summarization

NVIDIA Technologies Applied

The hexapod robot built on NVIDIA Jetson Orin Nano for real-time video preprocessing, H.264 encoding, and real time streaming to the Sarea Aegis platform. ROS 2 supports coordination between the robot, video pipeline, and inspection workflow, while NVIDIA Isaac ROS cuMotion and NVIDIA Isaac for Mobility support advanced robotic motion planning, perception, and localization. The Sarea Aegis platform is powered by EAR 100T with NVIDIA Jetson Thor and built on NVIDIA JetPack 7.1 platform.

 

Figure 3. Hexapod robot detection video stream

NVIDIA Metropolis and NVIDIA DeepStream SDK support the Sarea Aegis workflow for robot vision stream processing, object detection, heatmap generation, event analytics, and multi model inference.

NVIDIA TensorRT accelerates object detection through the Gst-nvinfer plugin, while NVIDIA Triton Inference Server supports scalable AI model serving for real time video analytics.

Figure 4. AI powered video analytics with object detection, heatmap generation, and event analytics

For event monitoring, ROI and line crossing rules are handled through Gst-nvdsanalytics. When a target object or event is detected, such as a person wearing a yellow reflective vest, the Sarea Aegis can trigger a real time alarm on the interface.

 

Heatmaps are generated by collecting bounding box metadata from DeepStream over time, then converting the spatial distribution into color mapped visualizations. This gives operators a clearer view of activity patterns across monitored areas.

Figure 5. Sarea Aegis semantic video search and video summarization

The semantic video search and summarization function was developed using NVIDIA Metropolis Video Search and Summarization Agent, combined with EverFocus designed workflows for inspection review, event search, and event handling. Users can enter natural language search terms, such as “safety vest,” and quickly locate relevant video content from the inspection workflow.

Visit EverFocus at COMPUTEX 2026

Visitors are welcome to visit the EverFocus booth to experience the live demos for robotic automation and smart factory inspection.

 

Date: June 2 to 5, 2026
Venue: Taipei Nangang Exhibition Center Hall 2, 1st Floor
Booth: Q0702