Sem 3

I found a journal Named ”International Journal of Computer Vision” which have an impact factor of 3.6. It contains major technical advances for broad general interest and presents details from science and engineering from the growing field of computer vision. Since we the track ”image- and video technology” in the media technology-program and parts from computer vision is including in the track I find it could be very interesting for media technology students.

In the same journal as above, I read the text “Detection and tracking of occluded people”. The text is mainly about compare different methods and technologies for examine how well the different methods ant techniques works for detecting and tracking people in crowded street scenes using image technology. The aim in the paper is to examine which technology and method that works best, when other people occur people. Except techniques that are good on detecting one person, they also propose a new strategy that makes use of the pattern that appears when people overlapping each other, and are otherwise uncommon. So they basically take a single-person detector with two different methods, and a joint detector (for couple of people) using one method, and examine how much the people could occur each other, in terms of percentage, before the detector can’t detect the people, or make “false positives”. They try to give some theory of how the methods work, but it hard to understand. Maybe because I have to little background knowledge, but also because they use references in a way that force the reader to read the references, to fully understand.  So the could first show some mathematical calculation, then write like “that combing with (reference), makes something”. Sometimes they even use the references directly, like “we used the method in (reference)”.

Something that I think is little confusing is that they have a single-person detector, using method 1 & 2, and a joint detector using method 1, and set a table with all the result from there measures. When they set like a competition, and bold the one with the best result, but in the competition they only compare the single-detector using method 1 and the joint detector using method 1, and bold the best results, just between those. The single-person detector using method 2 is not in theirs competition, even if the result is in the table. So they bold the best result, that most appears in the joint detector, but the single-person detector with method 2, have many times better result, but is not bolded. The whole things gets little confusing, and it would be better to just compare the different detectors that using the same method, without printing the result from others method, and thereby just compare single-person detector with joint detector.

1.Briefly explain to a first year student what theory is, and what theory is not.

If I tried to explain for a first year student what theory is, I would probably say that is something that is hard to define, and could means different things for different subject, or different people. But it’s somehow an attempt to explain how or why something works, and the parts in the theory must be relevant by logical reasoning. The question, “why is that part needed” must be answered. Things like data or references, is not theory, even it could be a foundation for a theory. If for example data is used for develop a theory, it’s also imported that the question “why is the data it important?” could be answered.

 2. Describe the major theory or theories that are used in your selected paper. Which theory type (see Table 2 in Gregor) can the theory or theories be characterized as?

I would argue for the theory type in my selected paper is mainly of type 1, of an analysis type, there the answer the question, which detector is best? It could also be considered as a explanation type (type 2 in the table), sense it answers the question “why is it best”? However, it’s not so much about how comes it’s the best detector, where they theirs argument is mixed with references, instead of fully explaining in the paper how it comes that a certain detector is better, in a technical way.   

3. Which are the benefits and limitations of using the selected theory or theories?

An analysis could be both easier to write and read. Its less background knowledge needed for the reader, to understand an analysis. But its only show “what is”, and a reader don’t really have to know why it is so. An explanation will give more knowledge to the reader, even if it could require more background knowledge.

References

Tang, S., Andriluka, M. & Schiele, B., 2013. Detection and Tracking of Occluded People. International Journal of Computer Vision, (February). Available at: http://link.springer.com/10.1007/s11263-013-0664-6 [Accessed November 22, 2013].

Gregor, S. (2006). The Nature of Theory in Information Systems. MIS Quarterly, 30(3), 611-642.

Sutton, R. I. & Staw, B. M. (1995). What Theory is Not. Administrative Science Quarterly, 40(3), 371-384