README.md

---
license: mit
task_categories:
- image-classification
dataset_info:
  features:
  - name: image
    dtype: image
  - name: label
    dtype:
      class_label:
        names:
          '0': '1'
          '1': '2'
          '2': '3'
          '3': '4'
          '4': '5'
          '5': '6'
  splits:
  - name: train
    num_bytes: 16680937.624
    num_examples: 14694
  - name: validation
    num_bytes: 3191950.1
    num_examples: 2100
  - name: test
    num_bytes: 5527485.6
    num_examples: 4200
  download_size: 24752623
  dataset_size: 25400373.324
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: validation
    path: data/validation-*
  - split: test
    path: data/test-*
tags:
- computer-vision
- image-classification
- synthetic-data
pretty_name: Geometric Shapes Dataset
size_categories:
- 1K<n<10K
---

# Dataset Card for Geometric Shapes Dataset

## Dataset Description

### Dataset Summary

The Geometric Shapes Dataset is a synthetic dataset containing images of various geometric shapes with superimposed random text. Each image features a polygon (or just text) on a randomly colored background, with a short string of random characters partially obscuring the shape. This dataset is designed for tasks such as shape classification, image recognition, and robustness testing of computer vision models.

### Supported Tasks and Leaderboards

- **Image Classification**: The primary task for this dataset is multi-class image classification, where the goal is to identify the type of shape (or lack thereof) in each image.

| Label        | Shape Name | Image                                                        |
|:--------------:|:------------:|:--------------------------------------------------------------:|
| 1            | None       | ![None](example/1_None.jpg "None")                           |
| 2            | Circle     | ![Circle](example/2_Circle.jpg "Circle")                     |
| 3            | Triangle   | ![Triangle](example/3_Triangle.jpg "Triangle")               |
| 4            | Square     | ![Square](example/4_Square.jpg "Square")                     |
| 5            | Pentagone  | ![Pentagone](example/5_Pentagone.jpg "Pentagone")            |
| 6            | Hexagone   | ![Hexagone](example/6_Hexagone.jpg "Hexagone")               |


### Data Instances

Each instance in the dataset consists of:
- An image (50x50 pixels, RGB)
- A label indicating the type of shape

### Data Fields

- `image`: A 50x50 pixel RGB image in numpy array format.
- `label`: A string indicating the shape type. The labels correspond to the following shapes:
  - "1": No shape, only random text on a colored background
  - "2": Circle-like shape (polygon with 100 sides)
  - "3": Triangle
  - "4": Square
  - "5": Pentagon

Each image contains:
1. A randomly colored background
2. The specified geometric shape (except for label "1") filled with a different random color
3. A short string (4 characters) of random alphanumeric text overlaid on top, partially obscuring the shape

Note: The "Circle" (label "2") is approximated by a 100-sided polygon, which appears circular at the given resolution.



### Data Splits

The dataset is split into train (70%), validation (10%), and test (20%) sets.

## Dataset Creation

This dataset was created to provide a simple, controlled environment for testing image classification models, particularly in scenarios where the primary subject (the geometric shape) is partially obscured by text.

### Source Data

#### Data Generation

The data is synthetically generated using the 'generate_geometric_shapes_dataset.py' of the project from the project https://github.com/0-ma/geometric-shape-detector. No external data sources were used.


### Annotations

#### Annotation process

The annotations (labels) are generated automatically during the image creation process.


### Personal and Sensitive Information

This dataset does not contain any personal or sensitive information.


### Other Known Limitations

- The dataset is limited to a small set of predefined shapes.
- The image resolution is fixed at 50x50 pixels.
- The text overlay is always present, which may not reflect all real-world scenarios.


### Licensing Information

This dataset is released under the MIT License.