Never before have I seen such a display of fatalism as in the West Quadrangle on Friday morning. The smell of fear hung over the assembled crowd like a ritualised corpse.
Okay, my apologies for the purple prose - there were certain expectations on the GMM exam, and I feel the review ought to live up to them. Really, however, I'm just going through the motions, because it was - as exams go - quite unsurprising, especially given our expectations, and doom-laden predictions. There was certainly no outrage, and there were no seizures.
Question 1, on the assessed exercise, was fairly similar to the example question, with descriptions of various half-toning techniques, and the Electronic Screen Printing and Floyd-Steinberg algorithms in particular. Unfortunately, this was followed up by a question on Root-Mean-Square error in the compared pre- and post-half-toning results. Although it appeared in the sample answer to the sample question, few people really knew what was meant. We are (the Scots amongst us, at least) the last generation to go through much of high school without any statistical training, and, as a result, few were able to apply previously unheard-of concepts and then draw inferences from them. Some marks for drawing the histograms, though, hopefully.
Question 2 (which I did not attempt) concerned Bézier curves, and was - by all accounts - quite fair. I steered clear because of the 8-mark question on numerical accuracy: even though I had read the essays on the topic, I didn't feel I could reproduce them to that standard, never having implemented a curve-drawing algorithm.
Question 3 was the joint question, combining image segmentation with composition. Certainly, the marks for histogram generation and analysis seemed fair enough, but the questions about the "limiting resolution" in the subsequent parts (dealing with a quarter reduction in the dimensions of an image) were... unfamiliar, to say the least.
Question 4 was for fans of Dr. Gauss. Gaussian, DoG and LoG filters did abound, with some added discussion of convolution. It was a very fair question, split into several parts, with quite-generous marks available for each.
So, it could have been a lot worse. Indeed, GMM is such a large course (compared to the other 10-credit modules) that they could easily have gone for the jugular. Praise be, then, that they didn't and it's over.
Cheers,
Derek.
|