Frequently Asked Questions

Questions

1. What do we do?
2. What can Deep Vision do for my company?
3. In which environments does Deep Vision's technology work?
4. What hardware and software does Deep Vision's technology require?
5. How is Deep Vision's technology affected by the size of the data?
6. How do I use Deep Vision's technology?
7. How were the sample videos created?

Q1: What do we do?

Deep Vision delivers high-end, real-time machine perception technology to its licensees and partners.

[Top]

Q2: What can Deep Vision do for my company?

Deep Vision's sensor exploitation technology provides perceptive capability in machines, computers, and devices. Designed for maximum interoperability, it can be integrated into virtually any system.

[Top]

Deep Vision's technology is constrained only by the nature and quality of the sensor data. If the sensor explotation module is used in a dark environment, a normal visible light sensor, like the ones found in a standard camera, would not be appropriate. In this situation, the module would be best attached to a low-light or infra-red sensor to achieve the best results.

The technology has been designed to be configurable, so that it can adapt to different environments. For example, the technology can easily be configured to work in dark or poorly lit environments. The same technology, without modification, can be configured to function in bright environments.

When it comes to environmental complexity, Deep Vision's technology is limited only by the sensor resolution and data size. Deep Vision's technology is able to describe a single object of interest as well as multiple objects with minimal delay. The decision of whether an object is of interest is determined by the task at hand.

[Top]

Deep Vision's technology has been designed to be both hardware and operating system independent. It does not require any particular processor, camera, or operating system. The only requirement is that the sensor and sensor data be appropriate for the task.

[Top]

Q5: How is Deep Vision's technology affected by the size of the data?

Deep Vision's technology is able to analyse the sensor data with such speed that any increase in sensor resolution only marginally effects this process. We encourage our clients to use sensors capable of high resolutions with our technology. The increase in accuracy with higher resolutions often outweighs any additional hardware costs.

[Top]

Deep Vision always works closely with their clients in order to develop a customised solution that best suits their needs and specifications. The custom solution is in the form of an independent, black-box module.

Deep Vision will work closely with their client to establish the optimal system specification to ensure proper integration with the client's system. A typical black-box system specification would answer such questions as:

• Sensor input connection - Analogue, Digital, and/or IEEE 1394 (Firewire)
• User interaction requirements - Mouse, Keyboard, and/or Touch Screen
• Feedback requirements - Analogue or Digital CRT/LCD display, audio, and/or haptic
• Networking requirements
• Power requirements - AC 115V/240V and/or Battery
• Performance metrics - benchmarks, performance targets, etc.
• Durability - shock, vibration and/or water resistant
• Weight limitations

The black-box system model allows for the best interoperability with the client's existing system design while, at the same time, protecting Deep Vision's intellectual property.

[Top]

Deep Vision takes great care in producing the sample videos to present the viewer with an experience that mirrors a live demonstration. Below is the abbrieviated process that is used to produce the videos:

1. Record the video. The recorded video is saved in raw format. This ensures that video depicts exactly what is presented on the sensor at the time of the recording. All the videos are recorded using a Basler A601f Firewire camera operating at 30 FPS (640x480@8bpp).
2. The recorded video is then loaded into Deep Vision's demonstration application(s).
3. Whilst the demonstration application is processing the raw video, a seperate application (recordMyDesktop) is used to record the desktop.
4. The recording is then post-processed using a non linear video editor (kdenlive) to add the opening and closing slides.
5. The final video is then transcoded using FFMPEG into MPEG and OGG Theora video formats. The MPEG video should be viewable by any popular media player. The OGG Theora video is playable by any Linux distribution. Microsoft Windows users will need to download and install the OGG codec.

[Top]