ECCV 2018 Daily - Thursday

Daily Tuesday 11 5 DreamVu Inc. has developed the world’s first omni-stereo camera hardware and software platform for unifying human and machine vision. We speak to COO Rohan Bhatial and CTO Rajat Aggarwal , who have brought the company to the ECCV 2018 Expo , to find out more. Rohan explains: “ We build 360-degree depth sensors. It’s a first of its kind solution where you use just a single camera and our innovative optics design to capture and give a 360- degree RGB-D output for robots and autonomous machines .” There are a number of reasons why this is so special. One is that they use a single sensor in the system. Unlike other systems that use multiple sensors to capture such a field of view, they use a single camera, which decreases the power and computation requirements. The power requirement is just for a single imager, which is less than 5 volts. Even a single sensor available in the market now has a 10- 20 volt power requirement, yet only captures a very limited field of view. This gives DreamVu a great advantage over other products used in the industry. In addition to these benefits, because the field of view is 360 degrees, there is no blind spot in the stereo capture, which means robots can better understand and navigate in a scene. There is also no minimum depth distance requirement for the sensor, so you can go as close as possible to the product and the robot can navigate between very close objects. The frame rate of the camera is also very high . Rohan says that the product they are selling right now does a great job for indoor environments. The basic concept of this camera lies in the field of computational photography. Rohan reveals that getting from concept to product was a major challenge. Designing it, getting it manufactured, and making a product out of it has taken a few years. They have spent a lot of time working with manufacturing partners all over the world to get everything just right. Rajat tells us about another challenge they encountered: “ You cannot capture 360 degrees using a single image sensor. You minimally need at least two sensors. We came up with the computational optics solution , because of which you can now capture 360 degrees, as well as from different viewpoints, which brings in the stereo information. That’s how you capture the depth .” Rohan and Rajat have already received some great feedback from you this week, including that they should reduce the size, and that the software development kit they are providing should have the software for the autonomous navigation and the SLAM , so that everybody can use it as a plug and play device to their robots and then they can implement all those applications. Finally, Rohan says they are using ECCV as a platform for a limited launch of the product. They would like to work with 20-30 university and industry partners who can use the product in their R&D and engineering units. If anybody wants to buy this product or find out more, please visit the DreamVu booth here at the ECCV Expo. Expo: DreamVu “a single sensor in the system”