Alyon-AI/UMA-Dataset_VLM-Engine-v1

---
license: apache-2.0
---

# UMA-IA/PYXIS-Engine-v1

## Authors
- **Youri LALAIN**, Engineering student at French Engineering School ECE
- **Lilian RAGE**, Engineering student at French Engineering School ECE

## Dataset Summary

The **UMA-IA/PYXIS-Engine-v1** is a specialized dataset designed for training vision-language models in the field of **aerospace and aeronautical engineering**. It consists of high-quality **images of aircraft engine components paired with detailed captions** identifying and describing the visible parts. This dataset enables models to learn to recognize and analyze various engine components, making it ideal for **fine-tuning vision-language models** for technical visual recognition and analysis tasks in the aerospace industry.

## Dataset Details

- **Splits**:
- **Train**: Complete dataset for model training
- **Columns**:
- `image`: The image file of an aircraft engine component or cross-section
- `caption`: Detailed description of visible components in the image
- `image_id`: Unique identifier for each image
- `cui`: Technical classification identifier

## Dataset Structure

The dataset's primary focus is on providing high-quality annotated images of aircraft engine components with detailed technical descriptions. Each entry contains:

1. An image showing aerospace engine components from various angles and cross-sections
2. Detailed captions identifying components such as:
- Soufflante (Fan)
- Aubes (Blades)
- Rotor
- Stator
- Compresseur (Compressor)
- And other critical engine components

## Example Entries

| image | caption | image_id |
|-------|---------|----------|
| [Engine Image] | Composants visibles: - Soufflante - Aubes de soufflante - Rotor de soufflante... | 001269777_896x598_c_mirror |
| [Engine Image] | Composants visibles: - Soufflante - Aubes - Rotor - Stator - Compresseur... | 001269777_896x598_c_original |
| [Engine Image] | Composants visibles: - Soufflante - Aubes - Rotor - Stator - Compresseur... | 001269777_896x598_c_segment1 |

## Applications

This dataset is particularly valuable for:
- Training vision models to recognize aerospace engine components
- Developing diagnostic tools for engine maintenance
- Creating educational resources for aerospace engineering
- Enhancing technical documentation with automatic component recognition
- Supporting quality control processes in engine manufacturing

## How to Use

You can load this dataset using the Hugging Face `datasets` library:

```python
from datasets import load_dataset
dataset = load_dataset("UMA-IA/PYXIS-Engine-v1")

# Access the first sample
print(dataset["train"][0]["caption"])

# Display an image (if in a notebook environment)
from PIL import Image
import matplotlib.pyplot as plt

img = dataset["train"][0]["image"]
plt.figure(figsize=(10, 8))
plt.imshow(img)
plt.axis('off')
plt.title(dataset["train"][0]["caption"])
plt.show()