Calculus 2 Bilkent

[Barton Ostby]

Imageacquisition, sampling and quantization. Spatial domain processing. Image enhancement. Texture analysis. Edge detection. Frequency domain processing. Color image processing. Mathematical morphology. Image segmentation and region representations. Statistical and structural scene descriptions. Applications. Credits: 3 units

This course provides an introduction to image analysis and computer vision. We will start with low-level vision (early processing) techniques such as binary image analysis, filtering, edge detection and texture analysis. Then, we will cover mid-level vision topics such as image segmentation and feature extraction in detail. Finally, we will do case studies on several applications such as image classification, object recognition, and deep learning.

Students taking this course need to have good background on high-level programming, data structures, linear algebra, and matrix calculus.

No prior knowledge of image processing or computer vision is assumed.

The midterm exam will be held at ? during ? on ?. The exam will cover all topics from the beginning of the semester until the end of ? chapter. You are allowed to bring only the lecture notes (slides) without any additional notes.

Although there is no restriction for the topic that you will select (as long as it is related with the course contents), the students should seek consent for their project topic. Since we want to minimise possible overlaps over project topics, two groups will not be allowed to work on a very similar topic; The project topics will be confirmed on a first-come-first-served basis.

Train and evaluate a deep learning model for a data set you will select. Here you may use the third-party codes, but you CANNOT select any data sets that was used to pre-train any of the deep learning models (e.g., you cannot use the ImageNet data set to conduct your experiments). In this option, you are expected to get the model trained on your data set and obtain reasonable test set performances. An important part of this second option, is to explore the effects of different (hyper)parameters on the performance of the model. You need to select a deep neural network and report its performance on your data set. You are expected to select a not-so-easy data set.

You must prepare and submit a final report for your project along with the developed codes/trained models as two separate files (a pdf file for the report and a single archive file (e.g., zip, tar, rar) for the code) by ? on ?. The reports are expected to be around 6 pages and must follow the IEEE two-column format as described in their templates. Try to follow the format as closely as possible. Both the content and the format will be subject to grading.

Each group will give a presentation on their project in class. Every group member should take a part in presentation. You will have approximately 10-12 minutes for your presentation; we will have a discussion period of 5 minutes after the presentation. I will let you know the exact duration after the add-drop period.

The presentation content, its format and layout, and the way that you present it will affect your grade. The interest that your presentation attracts from the audience will also affect your grade.

Prepare your slides neatly and properly. It should contain at most 12 slides with reasonable content (only present the most important and interesting parts). Do not copy and paste any text/equation/table from a paper (if necessary, type them). If you need to use a figure (or an image) of a paper, take it but give a credit to this paper (so that we can understand how much effort you have put in preparing your presentation).

3a8082e126