# GCPDS Computer Vision Python Kit
A comprehensive toolkit for computer vision and segmentation tasks, developed by the GCPDS Team. This package provides state-of-the-art tools for training, evaluating, and deploying segmentation models with support for various architectures, loss functions, and performance metrics.
## 🚀 Features
- **Segmentation Models**: Support for UNet and other popular architectures
- **Multiple Loss Functions**: DICE, Cross Entropy, Focal Loss, and Tversky Loss
- **Performance Evaluation**: Comprehensive metrics including Dice, Jaccard, Sensitivity, and Specificity
- **Training Pipeline**: Complete training workflow with validation and monitoring
- **Experiment Tracking**: Integration with Weights & Biases (wandb)
- **Mixed Precision Training**: Automatic Mixed Precision (AMP) support for faster training
- **Flexible Configuration**: YAML/JSON-based configuration system
- **Visualization Tools**: Built-in visualization utilities for model predictions
- **Memory Management**: Efficient memory handling and cleanup utilities
## 📋 Requirements
- Python >= 3.8
- PyTorch >= 2.0.0
- CUDA-compatible GPU (recommended)
## 🔧 Installation
### From PyPI (when available)
```bash
pip install gcpds-cv-pykit
```
### From Source
```bash
git clone https://github.com/UN-GCPDS/gcpds-cv-pykit.git
cd gcpds-cv-pykit
pip install -e .
```
### Development Installation
```bash
git clone https://github.com/UN-GCPDS/gcpds-cv-pykit.git
cd gcpds-cv-pykit
pip install -e ".[dev,docs]"
```
## 📦 Dependencies
### Core Dependencies
- `torch>=2.0.0` - Deep learning framework
- `torchvision>=0.15.0` - Computer vision utilities
- `numpy>=1.21.0` - Numerical computing
- `opencv-python>=4.6.0` - Image processing
- `matplotlib>=3.5.0` - Plotting and visualization
- `wandb>=0.15.0` - Experiment tracking
- `tqdm>=4.64.0` - Progress bars
- `Pillow>=9.0.0` - Image handling
- `scipy>=1.9.0` - Scientific computing
- `pandas>=1.4.0` - Data manipulation
### Optional Dependencies
- **Development**: `pytest`, `black`, `flake8`, `isort`
- **Documentation**: `sphinx`, `sphinx-rtd-theme`
## 🏗️ Project Structure
```
gcpds_cv_pykit/
├── baseline/
│ ├── trainers/ # Training utilities
│ ├── models/ # Model architectures
│ ├── losses/ # Loss functions
│ ├── dataloaders/ # Data loading utilities
│ └── performance_model.py # Model evaluation
├── crowd/ # Crowd-specific implementations
├── datasets/ # Dataset utilities
└── visuals/ # Visualization tools
```
## 🚀 Quick Start
### Basic Training Example
```python
from gcpds_cv_pykit.baseline.trainers import SegmentationModel_Trainer
from torch.utils.data import DataLoader
# Define your configuration
config = {
'Model': 'UNet',
'Backbone': 'resnet34',
'Number of classes': 2,
'Loss Function': 'DICE',
'Optimizer': 'Adam',
'Learning Rate': 0.001,
'Epochs': 100,
'Batch Size': 8,
'Input size': [3, 256, 256],
'AMP': True,
'Device': 'cuda'
}
# Initialize trainer
trainer = SegmentationModel_Trainer(
train_loader=train_dataloader,
valid_loader=valid_dataloader,
config=config
)
# Start training
trainer.start()
```
### Model Evaluation Example
```python
from gcpds_cv_pykit.baseline import PerformanceModels
# Evaluate trained model
evaluator = PerformanceModels(
model=trained_model,
test_dataset=test_dataloader,
config=config,
save_results=True
)
```
### Custom Loss Function
```python
from gcpds_cv_pykit.baseline.losses import DICELoss, FocalLoss
# DICE Loss
dice_loss = DICELoss(smooth=1.0, reduction='mean')
# Focal Loss
focal_loss = FocalLoss(alpha=0.25, gamma=2.0, reduction='mean')
```
## 📊 Supported Models
- **UNet**: Classic U-Net architecture with various backbone options
- Backbones: ResNet, EfficientNet, and more
- Pretrained weights support
- Customizable activation functions
## 🎯 Loss Functions
- **DICE Loss**: Optimized for segmentation tasks
- **Cross Entropy**: Standard classification loss
- **Focal Loss**: Addresses class imbalance
- **Tversky Loss**: Generalization of Dice loss
## 📈 Metrics
The toolkit provides comprehensive evaluation metrics:
- **Dice Coefficient**: Overlap-based similarity measure
- **Jaccard Index (IoU)**: Intersection over Union
- **Sensitivity (Recall)**: True positive rate
- **Specificity**: True negative rate
All metrics are calculated both globally and per-class.
## 🔧 Configuration
The toolkit uses dictionary-based configuration. Key parameters include:
```python
config = {
# Model Configuration
'Model': 'UNet',
'Backbone': 'resnet34',
'Pretrained': True,
'Number of classes': 2,
'Input size': [3, 256, 256],
# Training Configuration
'Loss Function': 'DICE',
'Optimizer': 'Adam',
'Learning Rate': 0.001,
'Epochs': 100,
'Batch Size': 8,
# Advanced Options
'AMP': True, # Automatic Mixed Precision
'Device': 'cuda',
'Wandb monitoring': ['api_key', 'project_name', 'run_name']
}
```
## 📊 Experiment Tracking
Integration with Weights & Biases for experiment tracking:
```python
config['Wandb monitoring'] = [
'your_wandb_api_key',
'your_project_name',
'experiment_name'
]
```
## 🎨 Visualization
Built-in visualization tools for:
- Training/validation curves
- Model predictions vs ground truth
- Metric evolution across epochs
- Sample predictions
## 🧪 Testing
Run the test suite:
```bash
# Run all tests
pytest
# Run with coverage
pytest --cov=gcpds_cv_pykit
# Run specific test file
pytest tests/test_models.py
```
## 📚 Documentation
Build documentation locally:
```bash
cd docs
make html
```
## 🤝 Contributing
We welcome contributions! Please follow these steps:
1. Fork the repository
2. Create a feature branch (`git checkout -b feature/amazing-feature`)
3. Commit your changes (`git commit -m 'Add amazing feature'`)
4. Push to the branch (`git push origin feature/amazing-feature`)
5. Open a Pull Request
### Development Setup
```bash
# Clone the repository
git clone https://github.com/UN-GCPDS/gcpds-cv-pykit.git
cd gcpds-cv-pykit
# Install in development mode
pip install -e ".[dev]"
# Run code formatting
black gcpds_cv_pykit/
isort gcpds_cv_pykit/
# Run linting
flake8 gcpds_cv_pykit/
```
## 📄 License
This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.
## 👥 Authors
- **GCPDS Team** - *Initial work* - [gcpds_man@unal.edu.co](mailto:gcpds_man@unal.edu.co)
## 🙏 Acknowledgments
- PyTorch team for the excellent deep learning framework
- The computer vision community for inspiration and best practices
- Contributors and users of this toolkit
## 📞 Support
- **Issues**: [GitHub Issues](https://github.com/UN-GCPDS/gcpds-cv-pykit/issues)
- **Documentation**: [Read the Docs](https://gcpds-cv-pykit.readthedocs.io/)
- **Email**: your-email@example.com
## 🔄 Changelog
### Version 0.1.0 (Alpha)
- Initial release
- Basic UNet implementation
- Core loss functions
- Training and evaluation pipeline
- Weights & Biases integration
---
**Note**: This project is in active development. APIs may change between versions. Please check the documentation for the latest updates.
Raw data
{
"_id": null,
"home_page": "https://github.com/UN-GCPDS/gcpds-cv-pykit",
"name": "gcpds-cv-pykit",
"maintainer": "GCPDS Team",
"docs_url": null,
"requires_python": ">=3.8",
"maintainer_email": "gcpds_man@unal.edu.co",
"keywords": "computer vision, segmentation, deep learning, pytorch, unet, medical imaging, image processing, machine learning, artificial intelligence",
"author": "GCPDS Team",
"author_email": "gcpds_man@unal.edu.co",
"download_url": "https://files.pythonhosted.org/packages/de/78/0b5cffaada6a77fba4e225b5bcf742d17de863c644a430a53bbcb7824a3b/gcpds-cv-pykit-0.1.1.tar.gz",
"platform": "any",
"description": "# GCPDS Computer Vision Python Kit\r\n\r\nA comprehensive toolkit for computer vision and segmentation tasks, developed by the GCPDS Team. This package provides state-of-the-art tools for training, evaluating, and deploying segmentation models with support for various architectures, loss functions, and performance metrics.\r\n\r\n## \ud83d\ude80 Features\r\n\r\n- **Segmentation Models**: Support for UNet and other popular architectures\r\n- **Multiple Loss Functions**: DICE, Cross Entropy, Focal Loss, and Tversky Loss\r\n- **Performance Evaluation**: Comprehensive metrics including Dice, Jaccard, Sensitivity, and Specificity\r\n- **Training Pipeline**: Complete training workflow with validation and monitoring\r\n- **Experiment Tracking**: Integration with Weights & Biases (wandb)\r\n- **Mixed Precision Training**: Automatic Mixed Precision (AMP) support for faster training\r\n- **Flexible Configuration**: YAML/JSON-based configuration system\r\n- **Visualization Tools**: Built-in visualization utilities for model predictions\r\n- **Memory Management**: Efficient memory handling and cleanup utilities\r\n\r\n## \ud83d\udccb Requirements\r\n\r\n- Python >= 3.8\r\n- PyTorch >= 2.0.0\r\n- CUDA-compatible GPU (recommended)\r\n\r\n## \ud83d\udd27 Installation\r\n\r\n### From PyPI (when available)\r\n```bash\r\npip install gcpds-cv-pykit\r\n```\r\n\r\n### From Source\r\n```bash\r\ngit clone https://github.com/UN-GCPDS/gcpds-cv-pykit.git\r\ncd gcpds-cv-pykit\r\npip install -e .\r\n```\r\n\r\n### Development Installation\r\n```bash\r\ngit clone https://github.com/UN-GCPDS/gcpds-cv-pykit.git\r\ncd gcpds-cv-pykit\r\npip install -e \".[dev,docs]\"\r\n```\r\n\r\n## \ud83d\udce6 Dependencies\r\n\r\n### Core Dependencies\r\n- `torch>=2.0.0` - Deep learning framework\r\n- `torchvision>=0.15.0` - Computer vision utilities\r\n- `numpy>=1.21.0` - Numerical computing\r\n- `opencv-python>=4.6.0` - Image processing\r\n- `matplotlib>=3.5.0` - Plotting and visualization\r\n- `wandb>=0.15.0` - Experiment tracking\r\n- `tqdm>=4.64.0` - Progress bars\r\n- `Pillow>=9.0.0` - Image handling\r\n- `scipy>=1.9.0` - Scientific computing\r\n- `pandas>=1.4.0` - Data manipulation\r\n\r\n### Optional Dependencies\r\n- **Development**: `pytest`, `black`, `flake8`, `isort`\r\n- **Documentation**: `sphinx`, `sphinx-rtd-theme`\r\n\r\n## \ud83c\udfd7\ufe0f Project Structure\r\n\r\n```\r\ngcpds_cv_pykit/\r\n\u251c\u2500\u2500 baseline/\r\n\u2502 \u251c\u2500\u2500 trainers/ # Training utilities\r\n\u2502 \u251c\u2500\u2500 models/ # Model architectures\r\n\u2502 \u251c\u2500\u2500 losses/ # Loss functions\r\n\u2502 \u251c\u2500\u2500 dataloaders/ # Data loading utilities\r\n\u2502 \u2514\u2500\u2500 performance_model.py # Model evaluation\r\n\u251c\u2500\u2500 crowd/ # Crowd-specific implementations\r\n\u251c\u2500\u2500 datasets/ # Dataset utilities\r\n\u2514\u2500\u2500 visuals/ # Visualization tools\r\n```\r\n\r\n## \ud83d\ude80 Quick Start\r\n\r\n### Basic Training Example\r\n\r\n```python\r\nfrom gcpds_cv_pykit.baseline.trainers import SegmentationModel_Trainer\r\nfrom torch.utils.data import DataLoader\r\n\r\n# Define your configuration\r\nconfig = {\r\n 'Model': 'UNet',\r\n 'Backbone': 'resnet34',\r\n 'Number of classes': 2,\r\n 'Loss Function': 'DICE',\r\n 'Optimizer': 'Adam',\r\n 'Learning Rate': 0.001,\r\n 'Epochs': 100,\r\n 'Batch Size': 8,\r\n 'Input size': [3, 256, 256],\r\n 'AMP': True,\r\n 'Device': 'cuda'\r\n}\r\n\r\n# Initialize trainer\r\ntrainer = SegmentationModel_Trainer(\r\n train_loader=train_dataloader,\r\n valid_loader=valid_dataloader,\r\n config=config\r\n)\r\n\r\n# Start training\r\ntrainer.start()\r\n```\r\n\r\n### Model Evaluation Example\r\n\r\n```python\r\nfrom gcpds_cv_pykit.baseline import PerformanceModels\r\n\r\n# Evaluate trained model\r\nevaluator = PerformanceModels(\r\n model=trained_model,\r\n test_dataset=test_dataloader,\r\n config=config,\r\n save_results=True\r\n)\r\n```\r\n\r\n### Custom Loss Function\r\n\r\n```python\r\nfrom gcpds_cv_pykit.baseline.losses import DICELoss, FocalLoss\r\n\r\n# DICE Loss\r\ndice_loss = DICELoss(smooth=1.0, reduction='mean')\r\n\r\n# Focal Loss\r\nfocal_loss = FocalLoss(alpha=0.25, gamma=2.0, reduction='mean')\r\n```\r\n\r\n## \ud83d\udcca Supported Models\r\n\r\n- **UNet**: Classic U-Net architecture with various backbone options\r\n - Backbones: ResNet, EfficientNet, and more\r\n - Pretrained weights support\r\n - Customizable activation functions\r\n\r\n## \ud83c\udfaf Loss Functions\r\n\r\n- **DICE Loss**: Optimized for segmentation tasks\r\n- **Cross Entropy**: Standard classification loss\r\n- **Focal Loss**: Addresses class imbalance\r\n- **Tversky Loss**: Generalization of Dice loss\r\n\r\n## \ud83d\udcc8 Metrics\r\n\r\nThe toolkit provides comprehensive evaluation metrics:\r\n\r\n- **Dice Coefficient**: Overlap-based similarity measure\r\n- **Jaccard Index (IoU)**: Intersection over Union\r\n- **Sensitivity (Recall)**: True positive rate\r\n- **Specificity**: True negative rate\r\n\r\nAll metrics are calculated both globally and per-class.\r\n\r\n## \ud83d\udd27 Configuration\r\n\r\nThe toolkit uses dictionary-based configuration. Key parameters include:\r\n\r\n```python\r\nconfig = {\r\n # Model Configuration\r\n 'Model': 'UNet',\r\n 'Backbone': 'resnet34',\r\n 'Pretrained': True,\r\n 'Number of classes': 2,\r\n 'Input size': [3, 256, 256],\r\n \r\n # Training Configuration\r\n 'Loss Function': 'DICE',\r\n 'Optimizer': 'Adam',\r\n 'Learning Rate': 0.001,\r\n 'Epochs': 100,\r\n 'Batch Size': 8,\r\n \r\n # Advanced Options\r\n 'AMP': True, # Automatic Mixed Precision\r\n 'Device': 'cuda',\r\n 'Wandb monitoring': ['api_key', 'project_name', 'run_name']\r\n}\r\n```\r\n\r\n## \ud83d\udcca Experiment Tracking\r\n\r\nIntegration with Weights & Biases for experiment tracking:\r\n\r\n```python\r\nconfig['Wandb monitoring'] = [\r\n 'your_wandb_api_key',\r\n 'your_project_name',\r\n 'experiment_name'\r\n]\r\n```\r\n\r\n## \ud83c\udfa8 Visualization\r\n\r\nBuilt-in visualization tools for:\r\n- Training/validation curves\r\n- Model predictions vs ground truth\r\n- Metric evolution across epochs\r\n- Sample predictions\r\n\r\n## \ud83e\uddea Testing\r\n\r\nRun the test suite:\r\n\r\n```bash\r\n# Run all tests\r\npytest\r\n\r\n# Run with coverage\r\npytest --cov=gcpds_cv_pykit\r\n\r\n# Run specific test file\r\npytest tests/test_models.py\r\n```\r\n\r\n## \ud83d\udcda Documentation\r\n\r\nBuild documentation locally:\r\n\r\n```bash\r\ncd docs\r\nmake html\r\n```\r\n\r\n## \ud83e\udd1d Contributing\r\n\r\nWe welcome contributions! Please follow these steps:\r\n\r\n1. Fork the repository\r\n2. Create a feature branch (`git checkout -b feature/amazing-feature`)\r\n3. Commit your changes (`git commit -m 'Add amazing feature'`)\r\n4. Push to the branch (`git push origin feature/amazing-feature`)\r\n5. Open a Pull Request\r\n\r\n### Development Setup\r\n\r\n```bash\r\n# Clone the repository\r\ngit clone https://github.com/UN-GCPDS/gcpds-cv-pykit.git\r\ncd gcpds-cv-pykit\r\n\r\n# Install in development mode\r\npip install -e \".[dev]\"\r\n\r\n# Run code formatting\r\nblack gcpds_cv_pykit/\r\nisort gcpds_cv_pykit/\r\n\r\n# Run linting\r\nflake8 gcpds_cv_pykit/\r\n```\r\n\r\n## \ud83d\udcc4 License\r\n\r\nThis project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.\r\n\r\n## \ud83d\udc65 Authors\r\n\r\n- **GCPDS Team** - *Initial work* - [gcpds_man@unal.edu.co](mailto:gcpds_man@unal.edu.co)\r\n\r\n## \ud83d\ude4f Acknowledgments\r\n\r\n- PyTorch team for the excellent deep learning framework\r\n- The computer vision community for inspiration and best practices\r\n- Contributors and users of this toolkit\r\n\r\n## \ud83d\udcde Support\r\n\r\n- **Issues**: [GitHub Issues](https://github.com/UN-GCPDS/gcpds-cv-pykit/issues)\r\n- **Documentation**: [Read the Docs](https://gcpds-cv-pykit.readthedocs.io/)\r\n- **Email**: your-email@example.com\r\n\r\n## \ud83d\udd04 Changelog\r\n\r\n### Version 0.1.0 (Alpha)\r\n- Initial release\r\n- Basic UNet implementation\r\n- Core loss functions\r\n- Training and evaluation pipeline\r\n- Weights & Biases integration\r\n\r\n---\r\n\r\n**Note**: This project is in active development. APIs may change between versions. Please check the documentation for the latest updates.\r\n",
"bugtrack_url": null,
"license": "MIT",
"summary": "A comprehensive toolkit for computer vision and segmentation tasks",
"version": "0.1.1",
"project_urls": {
"Bug Reports": "https://github.com/UN-GCPDS/gcpds-cv-pykit/issues",
"Documentation": "https://gcpds-cv-pykit.readthedocs.io/",
"Homepage": "https://github.com/UN-GCPDS/gcpds-cv-pykit",
"Source": "https://github.com/UN-GCPDS/gcpds-cv-pykit"
},
"split_keywords": [
"computer vision",
" segmentation",
" deep learning",
" pytorch",
" unet",
" medical imaging",
" image processing",
" machine learning",
" artificial intelligence"
],
"urls": [
{
"comment_text": null,
"digests": {
"blake2b_256": "732ce6a7b89011951cc18ec415c66b7a77a54809984489d319a7da9f0e43485e",
"md5": "549ebdeab1afc36226659e34815287b5",
"sha256": "12eae8269836b402c5fcaaa589bcc17e56bcef5c2f4fd1c03987583daddb3fc5"
},
"downloads": -1,
"filename": "gcpds_cv_pykit-0.1.1-py3-none-any.whl",
"has_sig": false,
"md5_digest": "549ebdeab1afc36226659e34815287b5",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.8",
"size": 26209,
"upload_time": "2025-07-11T22:59:52",
"upload_time_iso_8601": "2025-07-11T22:59:52.904001Z",
"url": "https://files.pythonhosted.org/packages/73/2c/e6a7b89011951cc18ec415c66b7a77a54809984489d319a7da9f0e43485e/gcpds_cv_pykit-0.1.1-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": null,
"digests": {
"blake2b_256": "de780b5cffaada6a77fba4e225b5bcf742d17de863c644a430a53bbcb7824a3b",
"md5": "c98cda85dde3b1a80f04879be098f212",
"sha256": "ab9ead8bb66b801dbf90fb171ea89042622dec0f840cc3bc185fba2f6b1dade4"
},
"downloads": -1,
"filename": "gcpds-cv-pykit-0.1.1.tar.gz",
"has_sig": false,
"md5_digest": "c98cda85dde3b1a80f04879be098f212",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.8",
"size": 25984,
"upload_time": "2025-07-11T22:59:54",
"upload_time_iso_8601": "2025-07-11T22:59:54.559987Z",
"url": "https://files.pythonhosted.org/packages/de/78/0b5cffaada6a77fba4e225b5bcf742d17de863c644a430a53bbcb7824a3b/gcpds-cv-pykit-0.1.1.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2025-07-11 22:59:54",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "UN-GCPDS",
"github_project": "gcpds-cv-pykit",
"travis_ci": false,
"coveralls": false,
"github_actions": false,
"lcname": "gcpds-cv-pykit"
}
JSON
