Speed Sign Recognition by Convolutional Neural Networks

This video show a Speed Sign Detection and Recognition vision application that is uses a Convolutional Neural Network. This application is fully trainable by a set of example images. Instead of using algorithm complexity we used training data to improve recognition reliability. The algorithm has a very parallel nature and thereby scales very well on the current popular CUDA enabled Grapics Processors from Nvidia. We mapped the algorithm to a Nvidia Geforce GTX 460. The GPU implementation can perfome realtime processing of 35 fps on a 720×1280 HD video stream. More invormation can be found on our website: parse.ele.tue.nl

3 thoughts on “Speed Sign Recognition by Convolutional Neural Networks

  1. Impressive ! can you give us more details or explanation , am working on a person recognition in videos but am still in the first phase and i need some orientation

Leave a Reply