---
license: apache-2.0
language:
- vi
pipeline_tag: image-to-text
library_name: transformers
tags:
- image-captioning
- vietnamese
- vision-encoder-decoder
- deit
- gpt2
---

# Vietnamese Image Captioning

Model image captioning tieng Viet duoc train tren Flickr8k va 10k anh MSCOCO da dich sang tieng Viet.

Model nhan mot anh dau vao va sinh mot cau chu thich bang tieng Viet.

## Architecture

- Encoder: `facebook/deit-base-distilled-patch16-224`
- Decoder: `NlpHUST/gpt2-vietnamese`
- Transformers class: `VisionEncoderDecoderModel`

## Usage

```python
from PIL import Image
import torch
from transformers import AutoImageProcessor, AutoTokenizer, VisionEncoderDecoderModel

model_id = "slyviee/vietnamese-image-captioning"
device = "cuda" if torch.cuda.is_available() else "cpu"

tokenizer = AutoTokenizer.from_pretrained(model_id, use_fast=True)
image_processor = AutoImageProcessor.from_pretrained("facebook/deit-base-distilled-patch16-224")
model = VisionEncoderDecoderModel.from_pretrained(model_id).to(device)
model.eval()

image = Image.open("image.jpg").convert("RGB")
pixel_values = image_processor(images=image, return_tensors="pt").pixel_values.to(device)

with torch.no_grad():
    output_ids = model.generate(pixel_values, max_new_tokens=40, num_beams=4)

caption = tokenizer.decode(output_ids[0], skip_special_tokens=True)
print(caption)
```