LS1046A FRWY
• Integrates dual and quad 64-bit Arm®
Cortex®-A72 cores
• Packet processing acceleration and
high-speed peripherals
• Maintains hardware and software
compatibility between devices
LEARN MOREu
OBJECT AND FACIAL RECOGNITION
FOR ACCESS CONTROL
3. RCNN: This is a traditional algorithm that applies CNN classifiers to
multiple parts of the image, with different aspects to get more accuracy in
prediction at the cost of speed, even for smaller objects in an image.
Both YOLO and Mobile-net SSD lend themselves very well to CNN acceleration
techniques. Choice of algorithm often depends on the application, e.g. widget-
quality detection is more likely to go with Mobile-net SSD or RCNN for higher
accuracy, whereas pedestrian detection may go with YOLO for faster response times.
Object size also plays a role, though, as shown by the chart below. At large sizes, SSD
looks to perform more like RCNN, but as the size of an object decreases, so does the
accuracy of SSD.
Choosing the right object detection algorithm for your application is critical,
especially when safety is on the line. Hopefully you now have a better understanding
of which path to take. And lucky for you, NXP solutions make it easy to employ each
one of them. Keep reading through this eBook to learn more!
■
AI
11
