Texas Instruments Simplifying Smart CamerasBy: John Honovich, Published on Dec 06, 2008
To date, smart cameras are more promise than reality. While smart cameras receive significant press and a number of products are available, adoption is still very low - certainly under 1% of all cameras.
Here are the key points:
- TI's offering will make it quicker and easier for new companies to offer video analytics
- TI's offering will help video analytics providers make more robust and effective video analytic solutions
- This could weaken the strategic positioning of incumbents like ObjectVideo andioimage [link no longer available].
- Expect to see others provide and promote similar offerings, including Axis.
- Expect the practical customer impact to be in 2010 - 2012.
Providing no-cost libraries to developers is a fundamental tactic of technology companies. The easier the company can make it to use their platform, the more their platform will be used. For instance, the more developers use Microsoft's .NET (the most common platform for IP video surveillance software), the more licenses for Windows are sold. Likewise, the more companies use VLIB, the easier and cheaper it is to develop for the Da Vinci chips [link no longer available], the more Da Vinci chips are sold.
Indeed, similar libraries for computer vision/video analytics exist for Intel chips/PCs with Intel Performance Primitives [link no longer available] and OpenCV being the most well known. However, TI's VLIB is the first royalty free library for cameras.
While TI's DaVinci chips are widely used, many of the largest camera companies develop their own chips in house (Axis being a key example). As such, this library will only work with companies that choose to use TI's chips in their cameras.
Quicker and Easier
Even if you have developed video analytics to run on PCs, a significant task in releasing smart cameras is making video analytics run well on cameras (i.e., porting the software from the PC to the camera). Different hardware architectures demand different optimizations.
TI's VLIB allows significant acceleration in time and effort for optimizing video analytics to run on cameras. This not only reduces the cost but opens up smart camera development to companies that would not otherwise have the in-house resources to build a smart camera. The end result is more companies offering products at lower costs.
More Robust Analytics
Problems with video analytic performance are common. Common issues such as false alerts are not merely an issue of poor software development. Such problems are routinely an issue of insufficient computing resources being available. While robust algorithms exist, it is often not possible or cost effective to deploy them because the computing power needed is either not available for chips inside of cameras or the cost of the chips are prohibitive.
By increasing the performance of the core functions in video analytics, this allows developers to employ more sophisticated approaches while staying withint the computing bounds of the chip.
Cheaper, more robust offerings is a threat against incumbents who have been independetly optimizing their offerings for years.
By giving away a resource that leading video analytic companies have been building for years, this makes it easier for other companies to enter and challenge. That being said, leaders such as ObjectVideo and ioimage have significant leads in number of deployments and overall system optimization. It will not be simple or quick to eliminate such leads.
We are likely to see a number of approaches that make video analytics simpler and more effective to deploy in cameras. For instance, Stretch, offers a different approach that automatically optimizes C/C++ code for use on chips inside cameras. Rather than require a new library for cameras, developers can use Stretch's SDK to rapidly port and optimize their code for cameras.
Axis frequently markets and stresses the importance of video analytics and the value of their partner network. It is possible that their recently released chip, the ARTPEC-3 [link no longer available], may provide a platform for 3rd party developers to run analytics.
Impact in 2010 to 2012
TI's release is likely to have an impact in the mid to long term. Because this is the release of a library, it will take time for companies to implement, release and widely deploy products based on this library. Nonetheless, it is an important signal and element that should be tracked in the growing maturity of smart cameras.