---

Image Analysis

---

Description

Repo for the Data Intensive MPhil 2024 Image Analysis coursework. Please see report/main.pdf for the full report.

Script outputs will be saved the ./outputs directory.

Installation

Set up

git clone git@github.com:VishalJ99/image_analysis.git

cd image_analysis

conda env create -f environment.yml

conda activate vj279_image_analysis

docker build -t vj279_image_analysis .

Usage

All scripts are designed to be run from the command line, the following are docker commands, if running outside docker make sure to activate the conda environment ia_vj279 first.

All outputs are saved to the outputs directory.

Running the scripts

Q1a - Lung CT segmentation

docker run -it -v $(pwd):/image_analysis vj279_image_analysis /bin/bash -c "source activate ia_vj279 && python src/q1a.py data/CT.png"

Q1b - Noisy Flower segmentation

docker run -it -v $(pwd):/image_analysis vj279_image_analysis /bin/bash -c "source activate ia_vj279 && python src/q1b.py data/noisy_flower.jpg"

Q1c - Coin segmentation

docker run -it -v $(pwd):/image_analysis vj279_image_analysis /bin/bash -c "source activate ia_vj279 && python src/q1c.py data/coins.png"

Q2a - Line Fitting

docker run -it -v $(pwd):/image_analysis vj279_image_analysis /bin/bash -c "source activate ia_vj279 && python src/q2a.py"

Q2b - Compressed Sensing Reconstruction

docker run -it -v $(pwd):/image_analysis vj279_image_analysis /bin/bash -c "source activate ia_vj279 && python src/q2b.py"

Q2c - Sparse Wavelet Reconstruction

docker run -it -v $(pwd):/image_analysis vj279_image_analysis /bin/bash -c "source activate ia_vj279 && python src/q2c.py data/river.jpg"

Q3a - Gradient Descent Convergence Rate

docker run -it -v $(pwd):/image_analysis vj279_image_analysis /bin/bash -c "source activate ia_vj279 && python src/q3a.py"    

Q3b - Learned Gradient Descent Optimisation

docker run -it -v $(pwd):/image_analysis vj279_image_analysis /bin/bash -c "source activate ia_vj279 && python src/q3b.py 10"

Note, docker version can not use GPU, for GPU acceleration, run the notebook src/coursework_LGD_filled_vj.ipynb on colab.

Contributing

License

Please see license.md

Author

Vishal Jain 2024-03-12