MVCNN training

Creating, training and evaluating a Multi-View Convolutional Neural Network. In particular, we will be checking how placing images belonging to incorrect classes among the instance images affects the overall performance of the model on unseen data.

Resources

• Dataset: ModelNet40v1
• Working environment: Ubuntu18.04 LTS / Python 3.6.9 / virtualenv
• Use the Makefile commands to:
  • create the project virtual environment
  • print the source terminal command to activate environment in terminal
  • run tensorboard to view training progress & results

Project structure

.
├── data                                # Stores the project data
│   └──modelnet40v1                     # Data folder to be sourced from the link provided in the resources
├── environment                         # Definition and contents of the project virtualenv
├── output                              # Stores model training results
│   └── lightning_logs                  # Generated automatically by pytorch-lightning during training
├── src                                 # Source files of the project
│   ├── dataset                         # Classes used in building and managing the project dataset
│   ├── model                           # Classes used in building and running the model
│   ├── settings                        # Data and model constants
│   ├── utilities                       # Utility functions used across the repo
│   ├── 01_data_analysis.ipynb          # Notebook describing the data used in the MVCNN model training
│   ├── 02_model_training.ipynb         # Notebook describing the assumptions, hyperparameters and the process of the MVCNN model training
│   ├── 03_model_evaluation.ipynb       # Notebook describing the evaluation of the training
│   └── model_training.py               # Script allowing model training from console
├── LICENSE.md                          
├── Makefile                            
└── README.md