My experiences with interdisciplinary teaching suggests
a related question is equally important:
What is/will be the role of computer vision for engineering
and science students in general?
The successful application of computer vision to real-world problems has resulted in a need for CV coursework for general engineering and science students, in addition to the needs of our own majors. Furthermore, the principles and techniques developed by the computer vision community are essential to the feasibility of incorporating scientific visualization and concept illustration into the classroom. Computer vision is becoming both a mainstream subject of study (education FOR computer vision), and an enabling technology for new educational paradigms (education WITH computer vision).
The push to teach computer vision to broader, more general audiences than ever before is a sign of the maturity of the field. Computer vision has grown from a research area to a source of powerful tools for industry and other disciplines. Certainly, computer vision is a staple of Computer-Integrated Manufacturing systems and quality control. The European Economic Community is expected to invest millions in using computer vision to automate seafood inspection on fishing boats. Likewise, there is a tradition of computer vision in medicine, where many exciting projects such as telesurgery currently have high visibility. An emerging interdisciplinary area is the use of computer vision in capturing and analysing scientific phenomena. At the Colorado School of Mines, I saw three different departments (none of which taught computer vision) in one year make use of computer vision techniques. The Engineering Division used sequences taken with a high speed camera to confirm new theories in fluid dynamics, the Mining Engineering department measured the amount of Frazil ice buildup on drill bits from binary images, and Geophysics performed subsurface modeling and pattern recognition from remote sensing data.
It is incumbent on us to train the engineers and scientists to use computer vision effectively. But how do we teach computer vision to non-majors? Or even to undergraduate non-majors? Without a lengthy list of pre-requisites? Or without watering down the material for our majors? Run them through the Khoros tutorials and hope for the best? Reaching the general engineering and scientific student population may necessitate a separate class, radically restructured from the traditional computer vision classes being taught for CS/EE majors. One potential format is the "tailored electives" approach in place at the Georgia Tech Computer Integrated Manufacturing Systems (CIMS) Program The program offers an interdisciplinary graduate minor to students in Engineering, Computing, and Management. For the CIMS program, how to present a spectrum of technical subjects to non-majors within a reasonable course load is a very real issue. Their solution is to have a department tailor several courses for non-majors. The tailored courses provide a rigorous four week introduction to the technical subject area (typically material that would be required as pre-requisites), followed by a six week presentation of the broad issues in that field. Some depth is sacrificed in order to accommodate the initial review of necessary background material, but non-majors are exposed to the fundamentals of a topic and how to apply what they have learned.
Regardless of the demands of teaching computer vision for a wider population, new opportunities are arising for teaching with computer vision. These opportunities stem from the pedagogical shift towards visualization technologies. For example, in an editorial for Computer Applications in Engineering Education, Dr. Magdy Iskander reports that the Conceptual Learning of Science Consortium is investigating the impact of visible images of objects and behaviors in reaching a conceptual and intuitive understanding of scientific concepts. Note that the use of visible images is expected to be a key component. Whereas scientific visualization or concept illustration does not directly involve computer vision, the implementation will. Already, concerns are being raised in the engineering education literature about how to present a sequence of images associated with a class topic, permit the students to interact and manipulate them, and store and/or compress the data. The advent of remote learning, with its incumbent transmission demands, will further exacerbate the challenges. Educators will rely on the computer vision community to provide enabling solutions to the image processing bottleneck.
The educational role of the computer vision community is expanding in new directions. Fortunately, at least two resources are available to facilitate the development of new, different courses and the integration of computer vision techniques into new pedagogical styles. First, forums already exist for sharing what instruction techniques have (and have not) worked. Below is list of five journals and conferences where articles on the use and integration of computer vision for engineering education routinely appear. Second, innovative educational development is fundable, by either government (e.g., NSF) or by industry (e.g., learning consortia). I know of at least one case where NSF sponsored the development of an undergraduate course in machine vision.
Unfortunately, neither of these resources are dedicated to computer vision, and so may not be as effective or as accessible as they could be. Therefore, I would like to see a workshop dedicated to all aspects of computer vision in education to be held at least every two years, possibly in conjunction with a vision conference, and sponsored in part by either government and/or industry. Such a workshop could help identify how to reuse or adapt what is already available for general education. For example, one of my colleagues has wished for years for "khoros-lite," a PC, disk space friendly version oriented towards demonstrations and teaching that our students could port to their home machines. The computer vision community would profit from the opportunity to work together, sharing our knowledge and experiences, to educate the next generation of engineers and scientists for and with computer vision!
EDUCATIONAL RESOURCES:
1. IEEE Transactions on Education, the journal for IEEE Education Society.
The Transactions provide short (3-4 pages) articles reporting on
course development and other topics of interest. The brevity of
the articles can be frustrating, but they typically provide an
excellent starting point for course development.
example articles:
"An Instructional Robotics and Machine Vision Laboratory,"
W.I. Clement and K.A. Knowles, vol. 37(1), Feb. 1994, pp 87-90.
"An Educational Image Processing/Machine Vision System," W.E. Richard, vol 37(1), Feb. 1994, pp 129-132.
"A Project Oriented Approach to the Teaching of Machine Perception," H.R. Myler, vol. 36(4), Nov. 1993, pp 380-383.
2. Annual Conference Proceedings of Frontiers in Education, sponsored by the IEEE Education Society.
The papers I've seen tend to include more of what didn't
work than the other publications, which I've found to be very useful!
Sessions are devoted to new technologies, such as computer-aided instruction
and algorithm visualization.
example article:
"A Computer Vision Course for Undergraduates- An Object-Oriented Approach,"
R.L. Place, 1993, pp 445-450.
3. Annual Conference Proceedings of the American Society for Engineering Education.
Large conference with sessions each year on DSP, the use of
multimedia, how to design scientific visualization laboratories, and other hot
topics.
4. Journal of Engineering Education, American Society for Engineering Education.
Same flavor as the annual conference.
5. Computer Applications in Engineering Education, John Wiley & Sons, pub.
This is a newer journal, which includes a diskette with software in each
issue.
example article:
"ProDim: Educational Software for Digital Image Processing,"
J.M. Ramirez, R.A. Garcia-Aguilar, D. Baez-Lopez, vol. 3(2), 1995, pp 165-172.