The Mission

Depthix AI brings visual intelligence to industrial automation by enabling real-time detection of failures in robotic workflows. Our platform analyzes live or recorded video from existing factory cameras to identify anomalies such as misgrips, misplacements and packaging defects. Unlike traditional rule-based monitoring, which reacts after failures occur, Depthix AI detects and localizes issues in real time, providing precise spatial coordinates and structured diagnostic data that downstream systems can act on immediately.

The Challenge

Modern industrial robots are monitored through PLC signals and sensor data, but these systems lack visual awareness. A robot may report a successful pick-and-place operation while the object is open, misaligned or dropped. Current monitoring is largely rule-based and reactive, identifying issues only after they have propagated downstream, leading to product defects, production downtime and costly manual inspection. While vision systems exist, they are often costly, difficult to integrate or limited to narrow inspection tasks, leaving a gap in real-time visual understanding of robot execution. As a result, failures are missed, pressure on human QA increases and overall efficiency and profitability are reduced.

The solution

Depthix AI is developing a video intelligence layer that integrates with existing robotic systems and requires no additional hardware, using cameras already installed on the factory floor to monitor robot operations. The platform combines vision-language AI models with classical computer vision to detect predefined failure cases such as misgrips, open boxes and incorrect placements. For each detected issue, it outputs precise bounding box coordinates, timestamps and confidence scores, enabling integration with robot control systems, quality pipelines and digital twin platforms. The approach is designed to be task-aware, constraint-driven and actionable, focusing on reducing false positives and providing outputs that systems can act on directly.