──What are the back-illuminated CAPD’s advantages and applications? *1 CAPD: Current Assisted Photonic Demodulator To realize this sensor, we developed a new back-illuminated CAPD*1 by combining CAPD technology, which is an IP of Sony Depthsensing Solutions, with Sony’s back-illuminated CMOS image sensor technology. There are various types of ToF, and we focused on developing an indirect-ToF (iToF) method that works by measuring the phase delay of returning light after it has been reflected by the target object. Yohtaro Yasu: With ToF (Time of Flight) image sensors, depth information is acquired by measuring the time it takes for an emitted light pulse to reflect and return to the sensor surface. ──Please tell us a little about ToF image sensors. We had to modify our original DSP plan over and over again before we were able to complete the intelligent image sensors. Also, research into AI models progresses at a tremendous speed, so it is necessary to adopt AI processing specifications that will continue to be relevant several years following commercialization. However, there were few products that used AI at the time, and it was unclear what kinds of features were expected. The core technology for processors specializing in AI was also being studied by a team from Sony Semiconductor Israel, and the original plan for the present DSP took form in 2017. As well as expertise in developing system LSI, Sony had other necessary technologies such as back-illuminated CMOS image sensors, stacked CMOS image sensors, and Cu-Cu direct bonding that allow for high quality image capturing and multiple simultaneous functions. ──What allowed Sony to develop the world’s first image sensors with AI processing functionality?Įki: The biggest reason is that Sony already had the foundations of the development in place when we started planning. In addition to making the development of tiny cameras with AI functionality possible, these image sensors are expected to enable various applications in the retail and industrial equipment industries, and combined with cloud computing, will contribute to realizing optimal systems. AI models can be rewritten and updated with the latest AI models to match various system environments and conditions. The reduced data usage of these image sensors will also help lower the power consumption of the cloud server. ──What are their advantages and applications?Įki: Extracting only the necessary data can reduce data transmission latency and power consumption of the camera and address privacy concerns when using cloud services. * According to Sony’s research (at the time of announcement) This configuration allows a single image sensor to handle everything from image capturing to AI processing without high-performance processors and external memory. The pixel chip is back-illuminated, and the logic chip has a number of features in addition to the conventional image sensor operation circuit, such as Sony’s original DSP (Digital Signal Processor) and memory to store the AI model of users’ choice. It realizes high-speed edge AI processing. Ryoji Eki: This is the world’s first* image sensor with AI processing functionality, and was announced in May 2020. ──Please tell us a little about intelligent vision sensors.
0 Comments
Leave a Reply. |