README.md

---
tags:
- chest-xray
- radiology
- report-generation
- mimic-cxr
license: apache-2.0
---

# LAPVQA — Radiology Report Generation (Frozen Off-the-shelf Encoders)

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

## Description

Generative decoder heads for **Radiology Report Generation (RRG)** on MIMIC-CXR,
trained on top of five **frozen** off-the-shelf vision encoders.
Each file contains the decoder weights; the encoder is not bundled.

## Results (MIMIC-CXR test set)

| Encoder | BLEU-4 | ROUGE-L | RadGraph-s |
|---|---|---|---|
| CLIP ViT-L/14 | 0.034 | 0.168 | 0.197 |
| Florence-2 | 0.035 | 0.169 | 0.205 |
| SigLIP | 0.036 | 0.168 | 0.211 |
| CoCa | 0.030 | 0.160 | 0.193 |
| OWLv2 | 0.034 | 0.169 | 0.197 |

SigLIP achieves the best RadGraph-s (0.211) among all frozen encoders, consistent with
its larger patch-token count (729 tokens at 384 px) providing finer spatial detail.

## Files

| File | Encoder backbone |
|---|---|
| `clip-vit-l14.pt` | CLIP ViT-L/14 |
| `siglip.pt` | SigLIP |
| `florence2.pt` | Florence-2 |
| `coca.pt` | CoCa |
| `owlv2.pt` | OWLv2 |