AdaptorNAS

This repository is the official implementation of AdaptorNAS: A New Perturbation-based Neural Architecture Search for Hyperspectral Image Segmentation.

An overview of AdaptorNAS:

Getting started

• Download or clone this repo to your computer.
• Run pip install -r requirements.txt to install the required Python packages.
• The code was developed and tested on Python 3.6.13 and Ubuntu 16.04.
• Note that this codebase runs on PyTorch Lightning library.

Preparing the UOW-HSI dataset

1. Download the data from https://documents.uow.edu.au/~phung/UOW-HSI.html.
2. Unzip the downloaded data into the desired location on your computer.
3. Copy the ./datasets/partition folder from this repo into the same location. This folder contains the image ids for each cross-validation fold.

Searching for the optimal decoder

The Python script for searching is the main_search.py. The script will search for the optimal decoder, then train the derived network from scratch.

Run python main_search.py --help for the full list of available arguments.

ResNet-34 encoder

An example of designing the optimal decoder for the ResNet-34 encoder with L=2, p=2, n=5.

python main_search.py --dataset_dir="path_to_the_data" --name="experiment_description" --nas_encoder="resnet34" --nas_layers=2 --nas_max_edges=2 --nas_selection_epochs=5 --batch_size=10 --gpu=1 --fold=1

MobileNet-V2 encoder

An example of designing the optimal decoder for the MobileNet-V2 encoder with L=2, p=2, n=5.

python main_search.py --dataset_dir="path_to_the_data" --name="experiment_description" --nas_encoder="mobilenet_v2" --nas_layers=2 --nas_max_edges=2 --nas_selection_epochs=5 --batch_size=11 --gpu=1 --fold=1 

EfficientNet-B2 encoder

An example of designing the optimal decoder for the EfficientNet-B2 encoder with L=3, p=2, n=5.

python main_search.py --dataset_dir="path_to_the_data" --name="experiment_description" --nas_encoder="efficientnet-b2" --nas_layers=3 --nas_max_edges=2 --nas_selection_epochs=5 --batch_size=8 --gpu=1 --fold=1