Designing a labeling application for image object detection

We seek to build a large collection of images with ground truth labels to be used for training object detection and recognition algorithms. Such data is useful for supervised learning and quantitative evaluation. To achieve this, we developed a user interface tool that allows easy image annotation. The tool provides functionalities such as drawing boxes, querying images, and browsing the database. Using this annotation tool, we can collect a large dataset that spans many object categories, often containing multiple instances over a wide variety of images. We quantify the contents of an existing dataset and compare against other state of the art datasets used for object recognition and detection. Also, we show how to extend our dataset to automatically enhance object labels with WordNet, discover object parts, and increase the number of labels using minimal user supervision