---
tags:
- chest-xray
- radiology
- visual-question-answering
- differential-vqa
- mimic-cxr
license: apache-2.0
---

# LAPVQA — Differential VQA (Captioning-Pretrained Encoder)

Part of the [LAPVQA collection](https://huggingface.co/collections/dmusingu/lapvqa).

## Description

DiffVQA head trained on the frozen **LAPVQA captioning-pretrained encoder**
([`lapvqa-pretrain-captioning`](https://huggingface.co/dmusingu/lapvqa-pretrain-captioning)).
Checkpoint is a plain `DiffVQAHead` state dict (vis_dim=1024).

## Results (test set)

| BLEU-4 | ROUGE-2 | RadGraph-s | BERTScore F1 |
|---|---|---|---|
| 0.468 | 0.562 | 0.303 | 0.938 |

## Loading

```python
import torch
from lapvqa.diffvqa.model import DiffVQAHead

ckpt = torch.load("pretrain-captioning_best.pt", map_location="cpu")
head = DiffVQAHead(vis_dim=1024)
head.load_state_dict(ckpt)
head.eval()
# pair with encoder_final.pt from lapvqa-pretrain-captioning
```