Vehicle Detection from Aerial Images using Deep Learning
Amrita University, Coimbatore
The work is aimed to produce a system that can detect vehicles from aerial images obtained from any airborne platform. The full potential of deep neural nets are used in the project. Its an ongoing work and on completion it will be fully automated vehicle detection solution.
Detecting Human Emotions : A Deep Learning Approach
Amrita University, Coimbatore
Automated detection of human facial expressions ranging across 6 emotional states was performed. The CNN model as of now gives an accuracy of just 56% (with significant overfitting), largely due to the small size of the synthetically generated facial images. Future work aims at incorporating traditional machine learning technologies along with CNN for classification task.
Neural Networks as Feature Extractors: An Experiment
Amrita University, Coimbatore
The Keras API was used with a Theano backend to train a standard Alexnet formulated by Alex Krizhevsky et.al, and use it as a feature extractor in place of standard machine learning feature extraction algorithms like SVD and HOG. The quality and relevance of these extracted feature maps of images were compared with prior methods by using them to train other classifiers like Random Forest, Support Vector Machines and AdaBoost.
â—„
1 / 1
â–º