# Satellite_4_Class-Image

**Repository Path**: hf-datasets/Satellite_4_Class-Image

## Basic Information

- **Project Name**: Satellite_4_Class-Image
- **Description**: Mirror of https://huggingface.co/datasets/LuminaAI/Satellite_4_Class-Image
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2025-08-30
- **Last Updated**: 2025-08-30

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

---
license: mit
---
## Satellite Imaging RCL Dataset

### Overview

This dataset contains satellite images structured explicitly for classification tasks using Lumina AI's Random Contrast Learning (RCL) algorithm via the PrismRCL application. Unlike LLM datasets, imaging datasets contain individual .png files organized by class.

### Dataset Structure

The dataset structure for image classification training:

```
satellite2-png/
    train/
        [class_1]/
            image_001.png
            image_002.png
            ...
        [class_2]/
            image_001.png
            image_002.png
            ...
    test/
        [class_1]/
            image_001.png
            image_002.png
            ...
        [class_2]/
            image_001.png
            image_002.png
            ...
```

- **Classes:** Folder names represent distinct image classes.
- **Images:** Each image file (.png) represents a single data sample.

### Image Data Preparation

For image datasets, PrismRCL has specific preparation requirements:

- Images must be in .png format.
- No resizing or normalization is required when using PrismRCL version 2.4.0 or later.
- File names must be unique across all class folders.

### Usage (Image-specific)

Use PrismRCL for training with image data:

```
C:\PrismRCL\PrismRCL.exe chisquared rclticks=10 boxdown=0 ^
  data=C:\path\to\satellite2-png\train testdata=C:\path\to\satellite2-png\test ^
  savemodel=C:\path\to\models\satellite_image_model.classify ^
  log=C:\path\to\log_files stopwhendone
```

### Explanation of Command

- **chisquared:** Specifies Chi-squared as the evaluation method for training.
- **rclticks:** Number of RCL iterations during training.
- **boxdown:** RCL-specific training parameter.
- **data & testdata:** Paths to training and testing image datasets.
- **savemodel:** Output path for the trained classification model.
- **log:** Directory for storing log files.
- **stopwhendone:** Automatically terminates the session after training completion.

### License

This dataset is licensed under the MIT License.

### Original Source

Prepared explicitly by Lumina AI for RCL-based image classification training. Please credit Lumina AI when using this dataset in research or applications.

### Additional Information

Refer to the PrismRCL Technical Documentation v2.6.2 for more detailed guidance on imaging data preparation and parameter specifications.