Image Analysis & Computer Vision – Tableware Classification Robot Project
A robot is being designed to clear a table and load a dishwasher. In order to accomplish this task, computer vision algorithms must be developed to classify the following:
An image database is provided.
Your project will be the development of the classification algorithm(s). You will work in groups of two. Note there are at least 10 images of each class. Use one-half of the images for training (developing) your algorithm, and the other images to test. Report your results on the test set.
The first task is to be able to identify the objects in the single object images. The second task (much harder) is to identify objects in the set images (multiple objects per image).
Approach: I suggest you divide the images into groups and develop a hierarchical classification scheme. For example, group the silverware together (knives, forks, spoons), the plates together (bowl, saucer/plate), the cup and glass together, and the napkins separately. Now we have four main classes, which we can initially indentify. Next, look within each class and develop another scheme to separate the items within a class. The first step in developing a classification algorithm is to find the best choice of preprocessing, segmentation (or edge detection), followed by morphological filtering. The next step is to label the image and select the object to be identified. Then, send the object to classification algorithm, and, hopefully, identify it correctly!