point-dl

Point based deep learning modules and associated tools for classification. There are Pytorch Geometric based models/tools as well as Pytorch based ones.

Contents

Pytorch Geometric Models

Model	Description	Reference
Classifier	A script used for classification using outputs of other models	NA
DGCNN	Dynamic Graph CNN model architecture	(Wang et al., 2019)
Dual Model	A script used to combine the outputs of two models based on defined method	NA
PointCNN	PointCNN model architecture	(Hell et al., 2022)
PointNet++	PointNet++ model architecture	(Qi et al., 2017)

Pytorch Geometric Utils

Util	Description
Augmentation	A script that performes augmentations on point clouds
Tools	A script with useful tools for point cloud deep learning and classification
Train	A script that defines the training/validation/testing process
Train Comp	A script that trains the models for composition

Pytorch Geometric Main Scripts

Example	Description
Comp Main	The main script that runs models for species composition
Dual Model Main	The main script to run a dual model for species composition
HP Optim	Script that tunes hyperparameters using Optuna

Pytorch Models

Model	Description	Reference
Point Augment	Adaptation of the Point Augment model (see below for contents)	(Li et al., 2020)
DGCNN	A Pytorch implementation of DGCNN	(Wang et al., 2019)
DGCNN Extended	An extended version of the Pytorch DGCNN model	NA

Pytorch Point Augment

Folder	File	Description
Augment	Augmentor	The augmentor (generator) model
Common	Loss Utils	The loss functions for the adapted model
Models	DGCNN	Pytorch implementation of DGCNN
Utils	Augmentation	A script that performes augmentations on point clouds
Utils	Tools	A script with useful tools for point cloud deep learning and classification
Utils	Train	A script that defines the training/validation/testing process
Point Augment	Main	The main scipt that runs the model

Pytorch Utils

Util	Description
Augmentation	A script that performes augmentations on point clouds
Tools	A script with useful tools for point cloud deep learning and classification
Train	A script that defines the training/validation/testing process

Pytorch Main Scripts

Example	Description
HP Tuner	Script that tunes hyperparameters using Optuna
Main	The main script that runs models

Tools

Script	Description
Create Labels	Creates the truth labels for every point cloud
Resample Point Clouds	Resamples point clouds based on different sampling methods