UWV_RAM

Reliability assessment of deep learning driven UWV

Installation

First of all, please set up a conda environment

conda create --name UWVRAM python==3.8
conda activate UWVRAM

This should be followed by installing software dependencies:

pip3 install matplotlib scikit-learn torch torchsummary torchvision tqdm imgaug tensorboard terminaltables

Fetch the Code

Fetch the source code for the reliabity assessment of UWV by run:

git clone https://github.com/Solitude-SAMR/UWV_RAM

Prepare the Dataset

Download the dataset and trained model weight from server using wget:

wget -P ./ https://cgi.csc.liv.ac.uk/~acps/datasets/SOLITUDE/data.zip

Unzip the folder and add to the root directory 'UWV_RAM/'.

Train the UWV Model and VAE Model

To train the yoloV3 model by yourself for UWV object detection, run

python -m pytorchyolo.train

To train the variantional autoencoder model by yourself to compress the UWV simulation images, run

python uwv_vae.py

Test the Reliability of Object Detection Model

run the following command to start testing:

python uwv.py

When the program is running, all test result for each demand (image) is saved to the output folder with format (latent_represenation, x_class, pmi). pmi is the abbreviation probabity of misclassification per input. Then you can visualize the robustness verification results of all the inputs, the update of reliability (pmi) by running

python plot.py