An Exploration of Deep Learning in Recognizing Household Objects

A realistic deep learning based mechanism for household object recognition is presented in this article. The deep learning encourages automatic feature extraction and works toward recognition under a stipulated workspace. Convolution Neural Network is one of the deep learning approaches which excelled in recognizing objects in images. It is implemented using 3 or more layers where each layer extracts one or more feature of the image. The Convolution Neural Network makes the user work easy by itself preprocessing the images fed by the user. The only thing here is to provide a large set of images to train or else augment the data. The model is going to compare the given image’s features with the features of the training set images in order to recognize image and classify that into the category it belongs among various category of household objects. Tensorflow framework has been used to develop this deep learning application. Parameter tuning and data augmentation has been employed to improve the performance of the system. This model is able to classify new instances of objects with an accuracy of 87.96% using dataset of 56 object categories.