This is the code repository for the paper Small Vessel Detection from Synthetic Aperture Radar (SAR) Imagery using Deep Learning .
This project was done as part of Stanford's CS 230, Deep Learning course.
Please see this link: https://cs230.stanford.edu/past-projects/#outstanding-projects for our posting.
Given the numerous models under consideration and the modular data downloading process, we present our code through interactive Jupyter notebooks.
Note that model weights, model output, and the dataset are not in this repo.
The original dataset can be found at: https://github.com/TianwenZhang0825/LS-SSDD-v1.0-OPEN
We make heavy use of Detectron2 which can be found at: https://github.com/facebookresearch/detectron2
The root directory features two notebooks training our best model and also performing inference.
final_model.ipynb is where we train the Improved model
final_evaluation.ipynb is where we perform inference on the Improved model
Contains our papers as part of the CS230 Deep Learning.
Collection of papers that we used in the course of this project
All of our training notebooks (which include baselines, experiments, and our final models presented in our paper)
Notebooks for preprocessing data and converting into Detectron2 format
Notebooks for generating plots for the writeup and developing the sea-land mask for copy-paste augmentation using Otsu's method
Notebooks for evaluating our models