Medical Visual Question Answering (VQA) Dataset for LoRA Fine-tuning PaliGemma

This dataset is designed for fine-tuning vision-language models, specifically PaliGemma, to perform medical visual question answering (VQA). It contains real-world medical images paired with patient-style questions and doctor-style answers.

---

Dataset Summary

• ✅ Multimodal: Images + Text
• ✅ Doctor-style professional answers
• ✅ Focused on common medical conditions
• ✅ Suitable for LoRA fine-tuning and VQA tasks

---

Dataset Structure

Each example contains:

---

Split Information

---

Example Usage

from datasets import load_dataset

dataset = load_dataset("siyunhe/medical-pilagemma-lora")

print(dataset["train"][0])
# Output:
# {
#   'image': <PIL.Image.Image>,
#   'question': 'Why is my skin peeling like this?',
#   'answer': 'This looks like dry skin or dermatitis, often caused by irritation or dehydration.'
# }

---

Intended Use

• LoRA fine-tuning for medical VQA models
• Medical Image Understanding
• Healthcare AI research
• Building Doctor-like multimodal assistants

---

Tags

• medical
• VQA
• vision-language
• healthcare
• LoRA
• fine-tuning
• PaliGemma
• multimodal

---

License

For research and educational purposes only.
Not for clinical or diagnostic use.

---

Citation

If you use this dataset, please cite:

@misc{medical_vqa_pilagemma,
  author = {Siyun He},
  title = {Medical Visual Question Answering Dataset for LoRA Fine-tuning PaliGemma},
  year = 2025,
  howpublished = {Hugging Face: siyunhe/medical-pilagemma-lora}
}