Efficient AI Platform for Image Processing
For the analysis of video streams using artificial intelligence (AI), normally extremely high performing hardware is required, which consumes a lot of energy. However, mother nature has proven to be an excellent teacher in this respect as well. Findings gained from brain research, which have led to the deciphering of processes associated with spatial vision, object recognition and movement identification in the cerebral cortex, are pointing the way to a method for highly efficient image processing. The “SlowFast” model is the result of emulating these processes on the basis of neural networks.
The implementation of this AI model has been analyzed at EDAG within the scope of research work whereby valuable findings have been gained for practical application. These help to better understand the influence exerted by various factors on the process of image recognition and thus enable targeted optimization of the relevant neural networks. The results are valid with respect to both the platform used, and as a means of improving different architectures of Convolutional Neural Networks (CNNs).
Cross-industry application possibilities
The SlowFast architecture is not only impressive in technical terms, but also in practice - with a wide range of applications:
- Medical: motion analysis for fall prevention, patient monitoring and operating theaters.
- Rail: Real-time detection for accident prevention on platforms and in maintenance areas.
- Industry: Safe human-robot collaboration and production control.
- Defense: Autonomous rescue systems to detect injured persons in the field.
- Retail: Discrete theft prevention and customer behavior analysis.
- Construction: Safety monitoring and protective measures on construction sites.
Download our white paper now, titled “Efficient Image Processing using AI”, where the research work is described in detail.