A computational camera attempts to digitally capture the essence of visual information by exploiting the synergistic combination of task-specific optics, illumination, sensors and processing. In this course we will study this emerging multi-disciplinary field at the intersection of signal processing, applied optics, computer graphics and vision, electronics, art, and online sharing through social networks. If novel cameras can be designed to sample light in radically new ways, then rich and useful forms of visual information may be recorded — beyond those present in traditional photographs. Furthermore, if computational process can be made aware of these novel imaging models, them the scene can be analyzed in higher dimensions and novel aesthetic renderings of the visual information can be synthesized.We will discuss and play with thermal cameras, multi-spectral cameras, high-speed, and 3D range-sensing cameras and camera arrays. We will learn about opportunities in scientific and medical imaging, mobile-phone based photography, camera for HCI and sensors mimicking animal eyes. We will learn about the complete camera pipeline. In several hands-on projects we will build physical imaging prototypes and understand how each stage of the imaging process can be manipulated.

This course provides an overview of robot mechanisms, dynamics, and intelligent controls. Topics include planar and spatial kinematics, and motion planning; mechanism design for manipulators and mobile robots, multi-rigid-body dynamics, 3D graphic simulation; control design, actuators, and sensors; wireless networking, task modeling, human-machine interface, and embedded software. Weekly laboratories provide experience with servo drives, real-time control, and embedded software. Students will design and fabricate working robotic systems in a group-based term project.

This course provides an overview of robot mechanisms, dynamics, and intelligent controls. Topics include planar and spatial kinematics, and motion planning; mechanism design for manipulators and mobile robots, multi-rigid-body dynamics, 3D graphic simulation; control design, actuators, and sensors; wireless networking, task modeling, human-machine interface, and embedded software. Weekly laboratories provide experience with servo drives, real-time control, and embedded software. Students will design and fabricate working robotic systems in a group-based term project.

MASLab (Mobile Autonomous System Laboratory) is a robotics contest. The contest takes place during MIT's Independent Activities Period and participants earn 6 units of P/F credit and 6 Engineering Design Points. Teams of three to four students have less than a month to build and program sophisticated robots which must explore an unknown playing field and perform a series of tasks. MASLab provides a significantly more difficult robotics problem than many other university-level robotics contests. Although students know the general size, shape, and color of the floors and walls, the students do not know the exact layout of the playing field. In addition, MASLab robots are completely autonomous, or in other words, the robots operate, calculate, and plan without human intervention. Finally, MASLab is one of the few robotics contests in the country to use a vision based robotics problem.

The course focuses on the problem of supervised learning within the framework of Statistical Learning Theory. It starts with a review of classical statistical techniques, including Regularization Theory in RKHS for multivariate function approximation from sparse data. Next, VC theory is discussed in detail and used to justify classification and regression techniques such as Regularization Networks and Support Vector Machines. Selected topics such as boosting, feature selection and multiclass classification will complete the theory part of the course. During the course we will examine applications of several learning techniques in areas such as computer vision, computer graphics, database search and time-series analysis and prediction. We will briefly discuss implications of learning theories for how the brain may learn from experience, focusing on the neurobiology of object recognition. We plan to emphasize hands-on applications and exercises, paralleling the rapidly increasing practical uses of the techniques described in the subject.

Most algorithms in computer vision and image analysis can be understood in terms of two important components: a representation and a modeling/estimation algorithm. The representation defines what information is important about the objects and is used to describe them. The modeling techniques extract the information from images to instantiate the representation for the particular objects present in the scene. In this seminar, we will discuss popular representations (such as contours, level sets, deformation fields) and useful methods that allow us to extract and manipulate image information, including manifold fitting, markov random fields, expectation maximization, clustering and others. For each concept -- a new representation or an estimation algorithm -- a lecture on the mathematical foundations of the concept will be followed by a discussion of two or three relevant research papers in computer vision, medical and biological imaging, that use the concept in different ways. We will aim to understand the fundamental techniques and to recognize situations in which these techniques promise to improve the quality of the analysis.

What are the circuits, mechanisms and representations that permit the recognition of a visual scene from just one glance? In this one-day seminar on Scene Understanding, speakers from a variety of disciplines -- neurophysiology, cognitive neuroscience, visual cognition, computational neuroscience and computer vision -- will address a range of topics related to scene recognition, including natural image categorization, contextual effects on object recognition, and the role of attention in scene understanding and visual art. The goal is to encourage exchanges between researchers of all fields of brain sciences in the burgeoning field of scene understanding.

An advanced seminar on issues of current interest in human and machine vision. Topics vary from year to year. Participants discuss current literature as well as their ongoing research.

Focuses on the problem of supervised learning from the perspective of modern statistical learning theory starting with the theory of multivariate function approximation from sparse data. Develops basic tools such as Regularization including Support Vector Machines for regression and classification. Derives generalization bounds using both stability and VC theory. Discusses topics such as boosting and feature selection. Examines applications in several areas: computer vision, computer graphics, text classification and bioinformatics. Final projects and hands-on applications and exercises are planned, paralleling the rapidly increasing practical uses of the techniques described in the subject.