AndresFelipeYule/AvesdelTolima

🦅 Aves del Tolima - High-Quality Bird Dataset

A comprehensive collection of 11,000 high-resolution images of 11 bird species from the Tolima region in Colombia. Perfect for machine learning, computer vision research, and biodiversity studies.

📊 Dataset Overview

Metric	Value
Total Images	11,000
Number of Species	11
Images per Species	1,000
Image Resolution	Minimum 600×600 pixels
Image Format	JPEG (Quality 95)
Total Size	~3.1 GB (uncompressed)
Unique Images	100% (0 duplicates)

🦜 Species Included

1. ATRAPAMOSCAS CARDENAL (Paroaria coronata)
2. BATARA CARCAJADA (Thamnophilus caerulescens)
3. BOBO RAYADO (Nystalus radiatus)
4. ELENIA MONTANA (Elaenia frantzii)
5. GALLITO DE ROCA (Rupicola peruvianus)
6. HORMIGUERO GUARDABOSQUE (Hypocnemis peruviana)
7. MARTIN PESCADOR GRANDE (Megaceryle torquata)
8. MARTIN PESCADOR VERDE (Chloroceryle amazona)
9. TIRANUELO CEJIAMARILLO (Phylloscartes ventralis)
10. TOROROI COMPADRE (Grallaria ruficapilla)
11. TUCANCITO ESMERALDA (Aulacorhynchus albivitta)

📥 How to Download

Option 1: Python (Recommended)

from datasets import load_dataset

# Load the dataset
dataset = load_dataset("AndresFelipeYule/AvesdelTolima")

# Access data
print(f"Total examples: {len(dataset['train'])}")

Install requirements:

pip install datasets huggingface-hub

Option 2: Direct Download

Click the "Download" button on this page.

Option 3: Git Clone

git clone https://huggingface.co/datasets/AndresFelipeYule/AvesdelTolima

🚀 Quick Start Examples

Basic Loading

from datasets import load_dataset

dataset = load_dataset("AndresFelipeYule/AvesdelTolima")
print(dataset)
print(f"Total images: {len(dataset['train'])}")

With PyTorch

from datasets import load_dataset
import torch
from torch.utils.data import DataLoader

dataset = load_dataset("AndresFelipeYule/AvesdelTolima")

def collate_fn(batch):
    images = torch.stack([torch.tensor(img['image']) for img in batch])
    return {'image': images}

loader = DataLoader(dataset['train'], batch_size=32, collate_fn=collate_fn)

With TensorFlow

import tensorflow as tf
from datasets import load_dataset

dataset = load_dataset("AndresFelipeYule/AvesdelTolima")

tf_dataset = dataset['train'].to_tf_dataset(
    columns=['image'],
    shuffle=True,
    batch_size=32
)

Memory-Efficient Streaming

# Stream without downloading entire dataset
dataset = load_dataset("AndresFelipeYule/AvesdelTolima", streaming=True)

for example in dataset['train']:
    print(example)
    break

💾 Dataset Structure

AvesdelTolima/
├── ATRAPAMOSCAS CARDENAL-PAROARIA CORONATA/
│   ├── ATRAPAMOSCAS CARDENAL-PAROARIA CORONATA_0001_xxxxx.jpg
│   ├── ATRAPAMOSCAS CARDENAL-PAROARIA CORONATA_0002_xxxxx.jpg
│   └── ... (1000 images)
├── BATARA CARCAJADA-THAMNOPHILUS CAERULESCENS/
└── ... (11 species total)

Naming Convention:

• Folders: COMMON_NAME-SCIENTIFIC_NAME
• Files: COMMON_NAME-SCIENTIFIC_NAME_XXXX_HASH.jpg
• Hash: First 10 characters of MD5 for deduplication

✅ Quality Assurance

All images have been validated:

• ✅ Minimum resolution: 600×600 pixels
• ✅ Minimum file size: 100 KB
• ✅ JPEG quality: 95
• ✅ No duplicates (MD5 hash verified)
• ✅ No corrupted files
• ✅ Sourced from trusted repositories

📊 Data Sources

Images collected from:

• GBIF - Global Biodiversity Information Facility
• iNaturalist - Research-grade observations
• Zenodo - Scientific repository
• Wikimedia Commons - CC-licensed media

All sources provide public or CC-licensed images.

🎯 Use Cases

1. Bird Classification - Train CNNs for species identification
2. Object Detection - YOLO, Faster R-CNN, RetinaNet
3. Feature Extraction - Transfer learning with ResNet, ViT
4. Biodiversity Research - Ecological studies
5. Conservation - Bird population monitoring
6. Computer Vision - General CV benchmarks

🔧 Advanced Usage

Save to Different Formats

# Parquet
dataset['train'].to_parquet("aves_dataset.parquet")

# CSV (metadata only)
dataset['train'].to_csv("aves_dataset.csv")

# Local directory
dataset['train'].save_to_disk("./aves_local")

Train/Test Split

from sklearn.model_selection import train_test_split

dataset = load_dataset("AndresFelipeYule/AvesdelTolima")
data = dataset['train']

train_idx, test_idx = train_test_split(
    range(len(data)), 
    test_size=0.2,
    random_state=42
)

train_set = data.select(train_idx)
test_set = data.select(test_idx)

📋 Troubleshooting

Issue	Solution
ModuleNotFoundError: datasets	pip install datasets
Slow download	Use streaming mode
Git LFS not installed	sudo apt-get install git-lfs
Low disk space	Use streaming or transfer learning

📚 Dataset Statistics

• Images per species: Exactly 1,000
• Total unique species: 11
• Geographic region: Tolima, Colombia
• Resolution range: 600×600 to 4096×4096
• Average file size: ~280 KB
• Compression ratio: 40% (3.1 GB → 1.2 GB)

📄 License & Attribution

This dataset combines images from multiple public sources:

• GBIF: Public domain/CC licenses
• iNaturalist: CC BY-NC (research use)
• Zenodo: CC licenses by authors
• Wikimedia Commons: CC BY-SA licenses

Please respect individual image licenses when using this dataset.

🏷️ Citation

If you use this dataset in research, please cite:

@dataset{aves_tolima_2025,
  title={Aves del Tolima: High-Quality Bird Image Dataset},
  author={Yule, Andres Felipe},
  year={2025},
  publisher={Hugging Face Datasets},
  url={https://huggingface.co/datasets/AndresFelipeYule/AvesdelTolima}
}

💬 Support

• Issues or questions? Check the dataset comments section
• Feature request? Leave feedback on the page
• Research use? Feel free to cite and use!

---

Made with ❤️ for biodiversity and machine learning research

Last updated: April 2025 | Dataset version: 1.0 | Status: ✅ Complete and public