Let cameras defeat AI fakes, says lidar maker

Let cameras defeat AI fakes, says lidar maker

Feature articles |
By Peter Clarke

Flash lidar vendor PreAct Technologies Inc. (Portland, Oregon) has said that advances in computer vision will enable cameras and computers to detect AI-generated content.

Amongst other things the lidar maker is proposing the use of depth information along with 2D visual information be aligned to blockchain technology to defeat deepfakes.

The company argues that this is a necessary development of computer vision as so-called ‘deepfake’ imagery is becoming common and making it difficult for human eyes to spot fakes. There was recent controversy over smart phones that use AI to enhance images beyond what is in the optical signal.

At the same time computer vision is poised to impact numerous industries including automotive, medical imaging, retail, health and safety, security, and entertainment.

PreAct reckons that cameras of the future will have access AI software to detect inconsistencies in AI-generated content such as micro-expressions on the face or biometrics. Computer vision has numerous lines of attack on deep fakes. These include: pixel analysis, metadata verification; source verification, the company said.

Depth and blockchain

Under source verification the company proposes the linking of images to blockchain or a similar decentralized database to allow the verification of an original source.

PreAct also proposes the capture of depth information along with an image. Depth maps would provide a 3D understanding of a scene that gives a signature to the captured event. This would make much more challenging for AI algorithms to generate deepfake images. Also, once the data is added to a blockchain network it becomes impossible to change without the network’s agreement.

The combination of depth map and blockchain would also de facto create a time stamp that would also help distinguish genuine images from forged content. Due to the decentralized nature of the blockchain, anyone can validate content against its timestamp, ensuring transparency and trustworthiness of the footage.

PreAct offers lidar for a wide range of industries including robotics, healthcare, logistics, security, industrial, consumer electronics, trucking, and automotive. It’s flash lidar produce 3D depth maps at sub-centimeter accuracy up to 20 meters distance.

Related links and articles:

News articles:

AI ‘camera’ uses internet data not light to form image

Intel Labs AI diffusion model generates 3D images from text

Ubicept raises US$8 million for SPAD-based imaging

SynSense raises US$10 million for smart vision sensor

Prophesee, SynSense form vision computing partnership

If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :    eeNews on Google News


Linked Articles