Breakthrough Slow-Motion Vision at a Fraction of the Cost and Power Consumption

AIStorm, in collaboration with Tower Semiconductor, has developed the Cheetah HS, a high-speed imager with first-layer AI capability, capturing up to 260,000 frames per second, 2,000 to 4,000 times faster than conventional CMOS sensors. This innovation combines ultra-high-speed imaging with charge-domain AI, significantly reducing system power requirements and bill-of-materials costs. The Cheetah HS is poised to revolutionize applications in robotics, drones, vibration and structure health monitoring, security and surveillance tracking, and more.

Key Takeaways:

• The Cheetah HS captures up to 260,000 frames per second, making it 2,000 to 4,000 times faster than conventional CMOS sensors.
• The imager's charge-domain architecture converts incoming photons to charge, computes the first neural-network layer in analog form, and outputs a pulse train for downstream processing.
• The Cheetah HS slashes system power requirements and bill-of-materials costs by 2-3 orders of magnitude compared to traditional high-speed cameras.
• The imager's proprietary charge-domain-based analog AI neurons enable efficient and low-power processing of high-speed events.
• The availability of Cheetah HS in both chip form and full reference-camera systems offers flexibility for designers to implement the technology in various applications.
• AIStorm's charge-domain processing technology has over 40 patents worldwide and allows for always-on sentry AI-in-sensor imaging solutions, high-speed imaging solutions, smart always-on solutions for audio applications, and human interface & biometric solutions.

Statistics:

• The Cheetah HS captures 260,000 frames per second.
• Conventional CMOS sensors are 2,000 to 4,000 times slower than the Cheetah HS.
• The imager reduces system power requirements by 2-3 orders of magnitude.
• AIStorm has over 40 patents worldwide in charge-domain processing technology.

Sources:

• AIStorm, press release, August 12, 2025
• Tower Semiconductor, press release, August 12, 2025