ComputationalPhotography-Course-Project

Computational Photography Course Project- Spring2023

Project Team

Luke Banaszak
Sakshi Maheshwari
Rick Suggs
Yitian Tang

Installing Dependencies

The project was created with Python 3.10 and has not been tested with any other version.

The Python dependencies can be installed by running the following command in the project directory.

pip install -r requirements.txt

Training PyTorch models can take a long time. If your computer is equipped with a GPU, that time could be reduced significantly. However, you may need to install different dependencies. See PyTorch - Get Started for instructions on installing GPU support on your computer.

For example, on Windows with Nvidia GPU, CUDA support can be enabled by first uninstalling Pytorch, then reinstalling with the compiled CUDA binaries:

pip uninstall torch torchvision
pip install torch torchvision --index-url https://download.pytorch.org/whl/cu117

Modules

Final_notebook.ipynb

The notebook begins with a simple set of cells to use the product's main functionality to generate a morphing GIF from a directory of jpg images.

The notebook then provides end to end examples of training the PyTorch models, all the way through generating a morph sequence from automatically predicted keypoints from multiple images.

morpher.py

Class definition that generates a morph sequence from a directory of images.

NeuralNet/train_cnn_resnet50.ipynb

Notebook covers results from training the pretrained Resnet 50 PyTorch model on the IMM Image Database Our trained model file for the IMM Image Database is larger than the Github file limit, but can be downloaded from here

NeuralNet/train_cnn_resnet50_ibug.ipynb

Notebook covers results from training the pretrained Resnet 50 PyTorch model on the IBug Image Database Our trained model file for the IBug Image Database is larger than the Github file limit, but can be downloaded from here

NeuralNet/cnn_utils.py

Module of utility functions used for training PyTorch models.

NeuralNet/face_classifier.py

Module of PyTorch Transforms and DataLoaders used for training PyTorch models.

NeuralNet/Models/resnet.py

Class definitions for pretrained Resnet 50 PyTorch models.

NeuralNet/Models/simpleModel.py

Class definitions for custom PyTorch models.

FaceMorphing/morphing.ipynb

Notebook covers manually defined correspondence points, creating morphing animations from multiple images, and experimenting with a population of images.

FaceMorphing/morphing_utils.py

Module of utility functions used for warping and morphing.

facial_keypoint_detection/1_load_and_visualize_data.ipynb

Notebook covers loading and visualizing data from the YouTube Faces Dataset

facial_keypoint_detection/2_define_the_network_architecture.ipynb

Notebook covers PyTorch CNN instantiation for YouTube Faces Dataset

facial_keypoint_detection/3_facial_keypoint_detection_complete_pipeline.ipynb

Notebook covers PyTorch CNN completed pipeline for YouTube Faces Dataset

facial_keypoint_detection/data_load.py

Module for YouTube Faces Dataset PyTorch DataLoader

facial_keypoint_detection/models.py

Module for YouTube Faces Dataset PyTorch Models

References

tbanas2 / morpher_generator Goto Github PK

morpher_generator's Introduction

ComputationalPhotography-Course-Project