A Japanese Tier 1 automobile manufacturing company.
Business Challenge
The client is entering in autonomous driving market. They wanted to develop a deep learning module that can recognize traffic speed limit signs in real time.
The algorithm had to be optimized on target platform for real time output.
Acclivis Contribution
CNN Layers library design and implementation
Dataset Collection and generation
Design and Implementation of CNN Net
Training and Testing of CNN model
CNN trained model generation
Product Features Developed/Supported by Acclivis
The application performs detection and classification of the traffic speed sign detection and classification. This application uses below processing modules to perform detection and classification:
Image Pre-processing module .
Deep Learning module
Classification module
Detection and extraction of the region of interest (i.e. Speed Sign Circle) is performed / implemented using image pre-processing modules, while Classification is performed using Deep learning module
Project Details
Language | C and C++
Initial development platform | NVIDIA TK1
DL Network | CNN VGG 16 NET
DL Network Layers | 16 Layer Network
Classifier | SoftMax
Number of Classes | 10
Deep Learning Library | Caffe
Client Benefit
Acclivis successfully delivered deep learning based speed sign which could be integrated in client's existing framework.
