The goal of an Automatic License Plate Recognition (ALPR) system is to capture and recognize a vehicle license plate. This is an important computer vision problem and has number of application domains: law enforcement, public safety agencies, and toll gate systems to name a few. At the heart of ALPR systems is the character recognition system as it is a unique identifier for any given vehicle. We construct an ALPR character recognition system by creating a dataset to simulate a captured license plate image, applying multiple binarization techniques to segment the characters from state, from the plate and from each other and finally using this dataset to train a convolutional neural network.
Currently the image recognition and classification implementing Convolutional Neural Networks is highly used, where one of the most important factors is the identification and extraction of characteristics, events, among other aspects; but in many situations this task is left only in charge of the neural network, without establish and apply a previous phase of image processing that facilitates the identification of patterns. This can cause errors at the time of image recognition, which in critical mission scenarios such as medical evaluations can be highly sensitive. The purpose of this paper is to implement a prediction model based on convolutional neural networks for geometric figures classification, applying a previous phase of color-space segmentation as image processing method to the test dataset. For this, it will be carried out the approach, development and testing of a scenario focused on the image acquisition, processing and recognition using an AR-Sandbox and data analysis tools. Finally, the results, conclusions and future works are presented.
Vehicle License Plate Detection and Recognition has become critical to traffic, security and surveillance applications. This contribution aims to implement and evaluate different techniques for License Plate Detection and Recognition in order to improve their accuracy. This work addresses various problems in detection such as adverse weather, illumination change and poor quality of captured images. After detecting the license plate location in an image the next challenge is to recognize each letter and digit. In this work three different approaches have been investigated to find which one performs best. Here, characters are classified through template matching, multi-class SVM, and convolutional neural network. The performance was measured empirically, with 36 classes each containing 400 images per class used for training and testing. For each algorithm empirical accuracy was assessed.
We present a reinforcement learning approach for detecting objects within an image. Our approach performs a step-wise deformation of a bounding box with the goal of tightly framing the object. It uses a hierarchical tree-like representation of predefined region candidates, which the agent can zoom in on. This reduces the number of region candidates that must be evaluated so that the agent can afford to compute new feature maps before each step to enhance detection quality. We compare an approach that is based purely on zoom actions with one that is extended by a second refinement stage to fine-tune the bounding box after each zoom step. We also improve the fitting ability by allowing for different aspect ratios of the bounding box. Finally, we propose different reward functions to lead to a better guidance of the agent while following its search trajectories. Experiments indicate that each of these extensions leads to more correct detections. The best performing approach comprises a zoom stage and a refinement stage, uses aspect-ratio modifying actions and is trained using a combination of three different reward metrics.
Automatic age and gender classification systems can play a vital role in a number of applications including a variety of recommendation systems, face recognition across age progression, and security applications. Current age and gender classifiers, are lacking crucial accuracy and reliability in order to be used in real world applications since most real-time systems have zero fault tolerant. This paper develops an end-to-end, deep architecture aiming to improve the accuracy and reliability of the age estimation task. We design a deep convolutional neural network (CNN) architecture for age estimation that builds upon a gender classification model. The system leverages a gender classifier to improve the accuracy of the age estimator. We investigate several architectures and techniques for the age estimator model with cross-modal learning, including an end-to-end model, using gender embedding of the input image, which leads to an increased accuracy. We evaluated our system on the Adience benchmark, which consists of real-world in-the-wild pictures of faces. We have shown that our system outperforms state-of-the-art age classifiers, such as [1] by 9%, by training a cross-modal age classifier.
