Digging Into the ShapeNetCore Dataset
In this article, we dive into the ShapeNetCore Dataset for the classification and segmentation of point cloud data and explore how to use it using Weights & Biases.
Created on December 16|Last edited on July 16
Comment
In this article, we cover the ShapeNetCore Dataset for the classification and segmentation of point cloud data. We also cover how to explore it using Weights & Biases.
Here's what we'll go over:
Table of Contents
Let's get started!
What Is ShapeNet?
ShapeNet is an ongoing effort to establish a richly-annotated, large-scale dataset of 3D shapes. ShapeNet provides global researchers with this data to enable research in computer graphics, computer vision, robotics, and other related disciplines.
The project is a collaborative effort between researchers at Princeton, Stanford, and TTIC.
The ShapeNetCore Subset
ShapeNetCore is a subset of the full ShapeNet dataset with single clean 3D models and manually verified category and alignment annotations. It covers 55 common object categories with about 51,300 unique 3D models. ShapeNetCore consists of 16 categories of objects which covers the 12 object categories of PASCAL 3D+, a popular computer vision 3D benchmark dataset.
Exploring the Dataset Using Weights & Biases
First, let's share a Weights & Biases Table that visualizes the ShapeNetCore dataset, and all the **16 object categories** that exist across `train-val` and `test` splits of the dataset. The table also shows the point clouds corresponding to each model along with the corresponding segmentation labels visualized in an interactive manner using the wandb.Object3D datatype.
Note: You can use the carousel below the visualization under the Point-Cloud column in the following table to scroll and view multiple point clouds.
If you wish to analyze the models in any particular category from the dataset, you can click on the eye icon corresponding to any run here to switch on/off the visualizations for that particular category 👇
16
Dataset Usage
Using ShapeNetCore in your ML worlflow using PyTorch Geometric's useful APIs
0
Leaderboards
Part Segmentation Leaderboard [Airplane]
3
Getting Started
Here are some reports that would help you get started with building your own machine-learning workflow on point cloud data...
Point Cloud Segmentation Using Dynamic Graph CNNs
In this article, we explore a simple point cloud segmentation pipeline using Dynamic Graph CNNs, implemented using PyTorch Geometric along with Weights & Biases.
Point Cloud Classification Using PyTorch Geometric
In this article, we explore how to classify point cloud data from 3D CAD Models, implementing the PointNet++ architecture and using PyTorch Geometric and W&B.
Add a comment
Iterate on AI agents and models faster. Try Weights & Biases today.