$ show zhixuhao/unet
This repository contains a Keras implementation of the U-Net architecture for biomedical image segmentation, specifically trained on the ISBI challenge membrane dataset. The code provides a straightforward pipeline: data loading and augmentation (via Keras ImageDataGenerator), model definition (U-Net with symmetric encoder-decoder), training script (main.py or notebook), and inference to generate predicted segmentation masks. The dataset is bundled within the repo (30 training images and masks, plus tests). The implementation relies on an outdated Keras 1.x API (deprecated `merge` function, `Model.fit_generator`) and has no dependency pinning or environment specification, making reproducibility low. While the functional components like the data generator and U-Net architecture are reusable patterns, the codebase is academically simple and exhibits significant technical debt relative to modern deep learning frameworks. It is best suited as a reference or educational starting point rather than production use.
The LLM identified 2 files worth highlighting for reuse:
The `adjustData` function handles image normalization and mask conversion to one-hot encoding for both binary and multi-class segmentation. It is self-contained, parameterized by `flag_multi_class` and `num_class`, and can be extracted and reused in other segmentation pipelines with minimal changes.
The `trainGenerator` function creates a paired data generator that applies identical augmentation (rotation, shift, shear, zoom, flip) to both images and masks using two synchronized Keras ImageDataGenerators. This pattern is a standard requirement for segmentation training and is cleanly abstracted, making it reusable for any folder-based image-mask dataset.