Text and image contrastive learning Python library implemented in C++

Installation

1. Clone the repository:

git clone https://github.com/eldarski/samaria.git
cd samaria

2. Build and run with Docker (recommended):

docker-compose up --build

Quick Usage Example

from samaria_contrast import ContrastiveLearningSystem

# Initialize
cls = ContrastiveLearningSystem()

# Test similarity between text and image
text = "赤い服の少女"  # Girl in red clothes
image_path = "data/images/1.jpeg"
similarity = cls.compute_similarity(text, image_path)
print(f"Similarity: {similarity:.3f}")

You can also run

python examples/simple_demo.py

Dependencies

• Python 3.9
• PyTorch
• OpenCV
• pybind11
• CMake

Model used

https://huggingface.co/rinna/japanese-clip-vit-b-16

---

https://exim.dev