Airbus ship image segmentation

Task

You are required to locate ships in images, and put an aligned bounding box segment around the ships you locate. Many images do not contain ships, and those that do may contain multiple ships. Ships within and across images may differ in size (sometimes significantly) and be located in open sea, at docks, marinas, etc.

For this metric, object segments cannot overlap. There were a small percentage of images in both the Train and Test set that had slight overlap of object segments when ships were directly next to each other. Any segments overlaps were removed by setting them to background (i.e., non-ship) encoding. Therefore, some images have a ground truth may be an aligned bounding box with some pixels removed from an edge of the segment. These small adjustments will have a minimal impact on scoring, since the scoring evaluates over increasing overlap thresholds.

Stack Technologies

• Python
• Keras
• Numpy, Pandas, Matplotlib
• Streamlit

Project structure

Source code contains code for exploratory data analysis, model training and interference. Model directoty contains pre-trained model and its weights. Also, there is app.py script is an entry point of the streamlit application. streamlit folder contains configuration for the application.

Project tree

• .streamlit
  • config.toml
• model
  • model.h5
  • weights.h5
• src
  • EDA
    • EDA.ipynb
  • inference
    • inference.py
    • utils.py
  • training
    • model.py
    • train.py
    • utils.py
  • app.py
  • constants.py
  • img_processing.py
• LICENSE
• README.md
• requirements.txt
• .gitignore
• .gitattributes

Installation

Clone project

git clone https://github.com/DHushchin/airbus-ship-detection

Create virtual environment

python -m venv venv

Activate it (depends on the OS)

venv\Scripts\activate

Install dependencies

pip install -r requirements.txt

Load dataset to /data directory

kaggle competitions download -c airbus-ship-detection

Run streamlit app locally

streamlit run src/app.py