https://github.com/aniketrajnish/3d-shapes-dataset-generator
Helps procedurally generate tailored 3D shape datasets
https://github.com/aniketrajnish/3d-shapes-dataset-generator
shapes-dataset shapes-generator unity3d
Last synced: 29 days ago
JSON representation
Helps procedurally generate tailored 3D shape datasets
- Host: GitHub
- URL: https://github.com/aniketrajnish/3d-shapes-dataset-generator
- Owner: aniketrajnish
- License: mit
- Created: 2023-02-23T13:20:11.000Z (about 2 years ago)
- Default Branch: main
- Last Pushed: 2023-12-28T00:23:53.000Z (over 1 year ago)
- Last Synced: 2024-05-21T06:50:38.091Z (about 1 year ago)
- Topics: shapes-dataset, shapes-generator, unity3d
- Language: ShaderLab
- Homepage:
- Size: 52.6 MB
- Stars: 15
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# 3D Shapes Dataset Generator
This tool is designed to help users create procedurally generated 3D shape datasets customized to their needs. It's built on top of my open source [Raymarching Engine](https://github.com/aniketrajnish/CS499-SDFNet/tree/main/Renderer) and runs over GPU. The engine supports over thirty primitives, three operations (Union, Intersection, and Subtraction), and varying color values (along with shadows).
A glimpse of the Final Dataset
Sample image with operations: Cone (Union) + Sphere (Union) + Box Frame (Subtraction)
## Getting Started
* Download the build file/windows installer from the [Releases](https://github.com/aniketrajnish/3D-Shapes-Dataset-Generator/releases/tag/1.0).
* If you need over the top features -
* Clone the repository
```
git clone https://github.com/aniketrajnish/3D-Shapes-Dataset-Generator.git
```
* Open and edit the source code (in the `/src` folder) in `Unity 2020.3.30f1` or later.
https://github.com/aniketrajnish/3D-Shapes-Dataset-Generator/assets/58925008/b6953128-a2d6-4023-b9ff-02ea1e3d8935
## Usage
* `Varying Camera Angle` (currently not functional): Set to `True` if you want to assign a different camera angle to each image in the dataset. Otherwise, the camera will look at the object keeping it in the center using `transform.LookAt()`.
* `Varying Orientation`: Set to `True` if you want to assign a unique random orientation (angle) to the individual shapes in the dataset. Otherwise, the shapes will be aligned with the axis using `Quaternion.identity`.
* `Varying Position`: Set to `True` if you want to assign a unique random position to the individual shapes in a cube of dimension 2 units centered at the origin. Otherwise, the shapes will be centered at the origin.
* `Randomize Shape Count`: Set to `True` if you want to randomize the number of shapes in each image between 0 and the `Max Shape Count`. Otherwise, every image will be generated with `Max Shape Count` number of shapes.
* `Max Shape Count`: Set the maximum number of shapes that each image in the dataset should have.
* `Dataset Size`: Set the number of images to be generated in the dataset.
* `Dataset Path`: Set the path where the dataset folder is to be created.
* `Resolution`: Set the width and height of the images (in pixels) to generate them accordingly.
* `Shapes and Operations`: These are enums that determine which shape index and operation index are to be taken into * consideration while generating each shape.
* `Seed`: Input a seed value to generate a dataset that has already been created before by assigning the seed value to the Random State.
Note that the `Varying Camera Angle` parameter is still under development and is not functional.
## Datasets
* The images are saved in the `../images` folder.
* The seed value of each random state is also exported in a txt file and can be used to
re-generate a dataset.
* These parameters are exported in the CSV sheet with all the image information as shown in table below.
| Column Name | Info |
| ------------- | ------------- |
| filename | Name of the image file |
| shape | Shape Index |
| operation | Operation Index |
| a,b,c,d,e,f,g,h,i,j,k,l | Dimensional parameters |
| hue, sat, val | HSV Values of the color |
| rot_x, rot_y, rot_z | Euler Angles |
| pos_x, pos_y, pos_z | Position Vector |
* Each row depicts information about a shape in the image of a dataset.
## Contributing
If you find a bug or have a feature request, please open an issue or submit a pull request.
## License
This project is licensed under the MIT License.