Scene Recognition System
The passion of developers who transcend the barriers of the workplace and debate from every possible angle in pursuit of an even better technology
The Scene Recognition System is a completely new technology. We have heard that the development process was fraught with various difficulties.
“To achieve an ‘even better technology’ was a challenge the Scene Recognition System absolutely had to overcome. There were plenty of difficulties just to do that. First of all, since the development process required us to undertake simulations showing how certain types of image data we photographed could produce certain types of output, we had to start by taking huge volumes of sample photos to be used in the simulations. To do that, we made a special camera that could use a 1,005-pixel RGB sensor to read the photo data. We connected the camera to a computer and took photos of lots of scenes we supposed would also be taken outdoors, although it looked a little strange to passers-by.”
This technology could have never come about if you weren't persistent, right? How did the data processing, one of the primary focuses of development, work out?
“A huge number of calculations is necessary to perform Scene Recognition, and the data processing inside the camera was extremely busy. The same information from the 1,005-pixel RGB sensor required different calculation processes depending on whether it was being analyzed for AE or AF. It took considerable effort to put together the required algorithms or to give the microprocessor enough memory to operate efficiently.”
Do the various calculations have to take place instantly?
“Yes. Even a single action requires extremely complex, high-speed calculations. Continuous shots in particular represent the hallmark of a successful camera. The camera has to work at maximum capacity when this is being done in AF. When the mirror is down, light is read by the sensor, and the instant the mirror goes up, calculation results are fed back, and when the shutter is pressed, the camera has to finish all the work. To repeat these operations instantaneously, calculations for various types of information to be passed on need to occur in an instant.”
I see. So, calculations are necessary for every movement of the camera?
“Exactly. It's extremely difficult to get the hardware and the data obtained in simulations to work together. We don't have much difficulty making simulations on the computer, but we had a very hard time getting the actual mechanism to work based on the simulation data. In 3D-Tracking, for instance, subject tracking information and AF need to link up and work together, but it didn't work at all at the beginning. We thought we had debugged the program and got the calculations perfect, too, but there were so many unknowns in this pioneering technology, it took a long while for us to make it work. We tested various possibilities in this process of trial and error, and once we finally got it to work in the camera, everyone involved in development and engineering felt a sense of achievement.”
You have been involved mainly in the development of metering technologies since you joined Nikon, so in the development of the Scene Recognition System, you had to expand into other fields such as AF, didn't you. I'm very interested in how you approached those challenges.
“Everyone in the development team came from completely different jobs, but our departments were nearby and we were in an environment where we could talk to each other quite easily. When talk of this development project came up, we had lots of discussions between departments. Especially when the idea was first proposed, it came under criticism from people wondering whether there was any sense in even attempting such a high hurdle. Once we put our heads together and talked openly about how we felt, I think we created an ideal environment for making something better.”
You knew it would be difficult, but you didn't say it was impossible. It is an admirable testimony to the interest and passion all the developers had for this new technology.
“I think it really was thanks to the efforts of so many people that we were able to do this. In the process of turning this new technology into a viable product, there really was a tremendous amount of ambiguity from the original idea to the final specifications we decided on. There were so many things we couldn't know just through experiments on the desk. In this area we had some members in the company who were in charge of making test photos and giving us constant advice. They were just like test drivers for new cars. We received lots of ideas from the hands-on experience of these people. Real ideas from the field are indeed very different from theoretical ideas, and they were indispensable to our development efforts.”
The harder the road we travel, the greater is our joy at the end. Just to hear someone say they were able to take a photo they could never take before.
How long did it take your team to develop the Scene Recognition System?
- Speaking of his future endeavors, Mr. Takeuch enthusiastically says: I want to explore applications to various camera models by considering their specifications and target users.
“About three years including time for the original idea.”
The D3 and D300 equipped with the Scene Recognition System have been released. Looking back on the development process, where did you get the energy to overcome all those difficulties?
“It was really our strong desire to succeed in creating this system. The difficulty in developing the system was the scale of the expectations people had for this technology. This provided the support we need to overcome obstacles in the development process. Of course, we have considerable satisfaction in having developed the technology, but more than that we look forward to the reaction of users. Though we developed these cameras with high-end, professional users in mind, professionals have the ability to make do with cameras with few functions. So if professionals use these cameras and say they are really useful, that would be a mark of success for us. The aim of the Scene Recognition System really has been to enable people to take photos they could never take before.”
We also look forward to future applications of the Scene Recognition System, which has a lot of potential.
“As a core technology, we would like to adapt it to future camera models, considering their specifications and target users. We want to start looking into various applications that will meet those needs.”