Perceptron, a startup co-founded by former Meta research scientists Armen Aghajanjyan and Akshat Shrivastava, has launched Isaac 0.5, a vision model designed to enable machines to “perceive, reason, and act” in industrial settings. The software aims to assist vision-guided robots in navigating complex environments such as warehouses and factory floors, as well as extracting visual intelligence from recorded videos. Isaac 0.5 is released as an open-weight model, allowing public inspection of its parameters and training materials.
The founders position their tool as distinct from existing models by emphasizing its general-purpose nature, designed to be flexible across various environments and tasks rather than being tailored for a single, repetitive function. Perceptron trained Isaac 0.5 on a million hours of video data, including general video, ego video (captured from a human’s perspective while performing tasks), and UMI video (recording repetitive human actions). The company claims to have internally built petabyte-scale, multi-modal datasets encompassing images, text, video, and robotic trajectories. Perceptron intends to market its software to a wide range of vendors, potentially integrating its intelligence layer across industries such as manufacturing, logistics, warehousing, security, mobility, media, and entertainment. Source