Thinci Inc. is a venture-backed, deep-learning vision processing start-up based in El Dorado Hills, California, with teams in California and Hyderabad, India. The company was founded by a highly skilled management team with years of experience in massively parallel processing architectures and the software structures to execute on these computing engines.
The company is currently in the final phase of producing its deep learning and vision processing solution comprising silicon and software that can be integrated into a wide range of applications, including advanced driver assistance systems in automotive; intelligent agents for personal electronics that enhance photos and video, explain the real world elements surrounding the user, protect the user from potential danger, and more; smart home automation systems that detect and prevent hazards, intelligently manages the energy the home consumes, and provides the optimum indoor climate.
In the realm of deep learning and machine processing, the human does not program the machine to perform a specific task. Instead, he describes a problem to the machine, provides it the tools for the solution, and supplies it the enormous amounts of data needed to derive an answer. The machine then intuits the best outcome, just as a human would. Thinci’s contribution to the deep learning and vision processing state-of-the-art is a silicon architecture capable of processing extremely large amounts of data in the same amount of time a human would, and the software infrastructure needed for the silicon to process this data.
The distinction that Thinci brings to the solutions currently coming onto the market is in the technique used to process the data. Deep learning and vision processing involves processing images, analogous to human vision. The image can be a picture or a large data base of “likes” collected on social media, collective purchases made on-line, fingerprints, etc. The human, for example, sees an image or collection of data and intuits a pattern—a person, place, or thing for the image; the movement of a stock or what’s trending on social media. Thinci’s deep learning and vision processing machine replicates this human function.
One approach to this processing is to apply graphics processors to compute image elements rather than render them as they do for video games. The approach Thinci has pioneered is to process the entire image in parallel, detecting a pattern the way the brain processes the image detected by the eye. This approach has the benefit of reducing the high frequency interaction with memory of graphics processors, with the benefit of reducing power consumption and boosting performance by eliminating the overhead of memory accesses.