Scratch-CNN

A modular CNN for detection and classification build with NumPy & Cython from scratch.

Setup

Create an environment with Conda:

conda env create --file standard_environment.yml --name scratch
conda activate scratch

Execute the sample pipeline on the test data:

python src/cnn_base.py

Build the fast_ops extension with:

python -m pip install -e .

If the extension is not built, the code falls back to the Python implementation. Right now only the max pooling function is supported.