Recognizingcharacter and digit from documents such as photographs which captured at astreet level is a very important factor in modern-day map making. For example,automatically identify an address accurately from street view images of thatbuilding. By using this information more precise map can be built and it canalso improve navigation services. Though normal character classification isalready a solved problem by computer vision but still recognizing digit orcharacter from the natural scene like photographs are still a harder problem. The reason behind the difficulties may be thenon-contrasting backgrounds, low resolution, blurred images, fonts variation,lighting etc.
Traditionalapproaches for classifying characters and digits from natural images wereseparated into two channels. First segmenting the images to extract isolatedcharacters and the perform recognition on extracted images. And this can bedone using multiple hand-crafted features *1 er ref and template matching. *1er refThemain purpose of this project is to recognize the street view house number byusing a deep convolutional neural network.
For this work, I considered the digit classification dataset of housenumbers which I extracted from street level images.http://ufldl.stanford.
edu/housenumbers/. This dataset is similar in flavor toMNIST dataset but with more labeled data. It has more than 600,000-digit imageswhich contain color information and various natural backgrounds and collectedfrom google street view images. To achieve the goal, I formed an applicationwhich will detect the number of just image pixels. Here, a convolutional neuralnetwork model with multiple layers is used to train the dataset and detect thehouse digit number with high accuracy. I used the traditional convolutionalarchitecture with different pooling methods and multistage features and finallygot 91.1% accuracy.