CS4243

Course Title

Computer Vision and Pattern Recognition

Grade

B+

Semester

AY23/24 S1

Review

The course is obviously about Computer Vision and Pattern Recognition. The first half is all about image processing, which is about how to manipulate images to extract certain features. We start off with the basics, like how to convert an image to grayscale, how to apply filters to an image, and how to detect edges in an image. We learn things like convolution, fourier transforms, and signal processing, all building to our intuition on how we can process the image. There is also motion detection, where we use two still images to determine the relative speed of an object in frame. These are all very basic concepts, but they are the building blocks for the more advanced topics later on.

Of course, as an AIML course, the fun doesn't stop there. We will now try to train the computer to extract and identify these features for us. Lots of image recognition use cases in here. But the part that doesn't quite sit right is the heavy emphasis on the really low-level AI concepts. Concepts like perceptrons and backpropagation are reiterated again in this course. But why though? Weren't they already covered in previous courses like CS2109S? It just felt like a good 3 weeks was wasted relearning the CS2109S concepts.

The workload was pretty low for the most part, around 3 take-home assignments with like one month to complete them. Even though they were lab assignments, we don't actually need to submit any code. Instead, we were given a Canvas MCQ quiz to input our answers. There was simply no room for error, which sucks because the MCQ questions are all very unclear and tricky. I probably lost a few marks there.

Other than that, there is tutorial attendance, where you only need to appear for 7/10 of the tutorial classes to get the full marks for attendance. A bit waste time, since the TA is only there for consultation and doesn't actually give any structured tutorials.

The final project involved creating an image classification model, which detects whether there is or isn't a weapon present in the image. To be honest, I was feeling a little down at this point in time, so I didn't contribute much to the project. At the start, yes, but nearing the end, I was just lost in every meeting. The final deliverable for the project is a poster and a presentation. The poster should be designed to be informative and you will be presenting based on the poster. No slides required.

The finals were full of tricky questions, and very unclear phrasing and instructions. It was troublesome, and probably why I ended up with a B+. Can't deal with unclear questions.