This repository contains solutions to the assignments of the course "Computer Vision" conducted by at the University of Salzburg, Department of Computer Science.
The course is part of the Master degree programs "Computer Science" and "Applied Image and Signal Processing". The Computer Vision course was conducted in WS 2017.
-
Babak Maser
https://github.com/BMaser -
Dominik Söllinger
https://github.com/dsoellinger
Warm up: Experiment with the skimage Harris corner detector, and explore the effect of changes in its parameters. E.g., study the effect of different values of Gaussian smoothing and different values of gamma in the response function.
Implementation of a very simple version of the Histogram-Of-Oriented-Gradients (HOG) descriptor of an image (region).
Implementation of a very simple image classifier based on HOG signatures. In particular, we will build a classifier for the infamous 15 scenes dataset. This dataset consists of (grayscale) scene images from 15 scenes.
Implementation of a convolutional neural network (CNN) for people counting (i.e., a regression task). Specifically, we use the UCSD Pedestrian dataset for this. This dataset has surveillance videos overlooking a sidewalk, where each frame is annotated by the number of people walking. For more details, see the dataset description available at the link above.
Implementation of our own neural network layer (in our case, an activation function) in PyTorch.
- Python 3
- Pytorch Framework
- Skimage Framework
- OpenCV Framework
- Numpy / Scipy / Pickle Framework