Geektime 2017
1 like306 views
The document discusses the challenges of deep learning, including the large computational power and costs required for graphic processors. It introduces Brodmann17's deep learning engine that enables visual recognition tasks on everyday devices at significantly lower costs than traditional approaches through its patented technique of training non-redundant networks. The company already provides deep learning software to large manufacturers, and sees the computer vision market growing substantially in the coming years.
Geektime 2017
- 1. Adi Pinhas | Co-founder & CEO
- 2. DID YOU ever ask yourself why they never show you the inside of self-driving cars?
- 3. DEEP LEARNING requires a massive amount of computational power
- 4. GRAPHIC PROCESSORS rule deep learning at a great cost of: Size Price Power Heat
- 5. CLOUD BASED deep learning created new challenges: Network Latency Security Privacy Complexity Scalability
- 6. DEEP LEARNING ENGINE for visual recognition tasks on everyday devices
- 7. WE ALREADY DELIVER deep learning software to large global device manufacturers Computer vision market is growing from $12B today to $30B+ in 2020
- 8. Brodmann17’s deep learning has better accuracy WITH SIGNIFICANTLY LOWER COSTS
- 9. Current approach large redundant network training compress, squeeze, approximate Out patented tech no redundancy network training