Crop Anomaly Detection with Semantic Segmentation

In this project I implement a homemade version of U-Net to identify anomalous regions in aerial photos of cropland. My neural net is capable of classifying anomalies into one of nine classes and was trained/tested on the 2021 Agriculture Vision dataset.

You can read the paper here!

• Skills: Data processing, computer vision, deep learning, model optimization, CUDA/GPU programming
• Tools: PyTorch, cuDNN, Pillow, NumPy