Add train/test split structure with README

README.md

@@ -0,0 +1,98 @@
+---
+license: mit
+task_categories:
+- image-classification
+tags:
+- animals
+- computer-vision
+- image-classification
+size_categories:
+- n<1K
+---
+
+# Animals Dataset
+
+## Dataset Description
+
+This dataset contains images of three animal categories: cats, dogs, and pandas.
+
+### Dataset Structure
+
+The dataset is organized into training and testing splits:
+
+```
+Animals_dataset/
+├── train/
+│   ├── cats/
+│   ├── dogs/
+│   └── panda/
+└── test/
+    ├── cats/
+    ├── dogs/
+    └── panda/
+```
+
+### Dataset Statistics
+
+- **Total Images**: 600
+- **Training Images**: 480 (80.0%)
+- **Testing Images**: 120 (20.0%)
+
+#### Class Distribution
+
+**Training Set:**
+- Cats: 160 images
+- Dogs: 160 images
+- Panda: 160 images
+
+**Testing Set:**
+- Cats: 40 images
+- Dogs: 40 images
+- Panda: 40 images
+
+### Usage
+
+You can load this dataset using the Hugging Face `datasets` library:
+
+```python
+from datasets import load_dataset
+
+# Load the entire dataset
+dataset = load_dataset("Melisa13/Animals_dataset")
+
+# Access train and test splits
+train_data = dataset['train']
+test_data = dataset['test']
+```
+
+Or using custom code:
+
+```python
+from huggingface_hub import hf_hub_download
+from PIL import Image
+import os
+
+# Download a specific file
+file_path = hf_hub_download(
+    repo_id="Melisa13/Animals_dataset",
+    filename="train/cats/cats_00001.jpg",
+    repo_type="dataset"
+)
+
+# Load image
+image = Image.open(file_path)
+```
+
+### Dataset Creation
+
+This dataset was split using scikit-learn's `train_test_split` with:
+- Test size: 20.0%
+- Random seed: 42
+
+### License
+
+MIT License
+
+### Citation
+
+If you use this dataset, please cite it appropriately.