Object Recognition with Improved Features Extracted from Deep Convolution Networks

Object recognition is a process for identifying a specific object in an image. Object recognition algorithms depends on matching, learning, or pattern recognition algorithms using appearance- based or feature - based techniques. Object recognition techniques include feature extraction and machine learning models, deep learning models such as CNN. Deep learning with convolution neural networks (CNN) has been proved to be very effective in feature extraction. CNN is comprised of one or more convolutional layers and then followed by one or more fully connected layers. In Image classifications, the work with classifiers aims at exploring the most appropriate classifiers for high level deep features. The features extracted from the image play an important role in image classification. Feature extraction is the process of retrieving the important data from the raw data. Feature extraction is finding the set of parameters that recognize the object precisely and uniquely. In feature extraction, each character is represented by a feature vector, which becomes its identity. The major goal of feature extraction is to extract a set of features, which maximize the recognition rate. In this work, the features extracted from CNN applied as input to train machine learning classifier and perform image classification. A systematic comparison between various classifiers is made for object recognition.