|
--- |
|
library_name: transformers |
|
datasets: |
|
- laicsiifes/flickr30k-pt-br |
|
language: |
|
- pt |
|
metrics: |
|
- bleu |
|
- rouge |
|
- meteor |
|
- bertscore |
|
base_model: |
|
- microsoft/swin-base-patch4-window7-224 |
|
pipeline_tag: text-generation |
|
--- |
|
|
|
# 🎉 Swin-DistilBERTimbau for Image Captioning |
|
|
|
Swin-DistilBERTimbau model trained for image captioning on [Flickr30K Portuguese](https://huggingface.co./datasets/laicsiifes/flickr30k-pt-br) (translated version using Google Translator API) |
|
at resolution 224x224 and max sequence length of 512 tokens. |
|
|
|
|
|
## 🤖 Model Description |
|
|
|
The Swin-DistilBERTimbau is a type of Vision Encoder Decoder which leverage the checkpoints of the [Swin Transformer](https://huggingface.co./microsoft/swin-base-patch4-window7-224) |
|
as encoder and the checkpoints of the [DistilBERTimbau](https://huggingface.co./adalbertojunior/distilbert-portuguese-cased) as decoder. |
|
The encoder checkpoints come from Swin Trasnformer version pre-trained on ImageNet-1k at resolution 224x224. |
|
|
|
The code used for training and evaluation is available at: https://github.com/laicsiifes/ved-transformer-caption-ptbr. In this work, Swin-DistilBERTimbau |
|
was trained together with its buddy [Swin-GPorTuguese](https://huggingface.co./laicsiifes/swin-gpt2-flickr30k-pt-br). |
|
|
|
Other models evaluated didn't achieve performance as high as Swin-DistilBERTimbau and Swin-GPorTuguese, namely: DeiT-BERTimbau, |
|
DeiT-DistilBERTimbau, DeiT-GPorTuguese, Swin-BERTimbau, ViT-BERTimbau, ViT-DistilBERTimbau and ViT-GPorTuguese. |
|
|
|
## 🧑💻 How to Get Started with the Model |
|
|
|
Use the code below to get started with the model. |
|
|
|
```python |
|
import requests |
|
from PIL import Image |
|
|
|
from transformers import AutoTokenizer, ViTImageProcessor, VisionEncoderDecoderModel |
|
|
|
# load a fine-tuned image captioning model and corresponding tokenizer and image processor |
|
model = VisionEncoderDecoderModel.from_pretrained("laicsiifes/swin-distilbert-flickr30k-pt-br") |
|
tokenizer = GPT2TokenizerFast.from_pretrained("laicsiifes/swin-distilbert-flickr30k-pt-br") |
|
image_processor = ViTImageProcessor.from_pretrained("laicsiifes/swin-distilbert-flickr30k-pt-br") |
|
|
|
# perform inference on an image |
|
url = "http://images.cocodataset.org/val2017/000000039769.jpg" |
|
image = Image.open(requests.get(url, stream=True).raw) |
|
pixel_values = image_processor(image, return_tensors="pt").pixel_values |
|
|
|
# generate caption |
|
generated_ids = model.generate(pixel_values) |
|
generated_text = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0] |
|
print(generated_text) |
|
``` |
|
|
|
## 📈 Results |
|
|
|
The evaluation metrics Cider-D, BLEU@4, ROUGE-L, METEOR and BERTScore are abbreviated as C, B@4, RL, M and BS, respectively. |
|
|
|
|Model|Training|Evaluation|C|B@4|RL|M|BS| |
|
|-----|--------|----------|-------|------|-------|------|---------| |
|
|Swin-DistilBERTimbau|Flickr30K Portuguese|Flickr30K Portuguese|66.73|24.65|39.98|44.71|72.30| |
|
|Swin-GPorTuguese|Flickr30K Portuguese|Flickr30K Portuguese|64.71|23.15|39.39|44.36|71.70| |
|
|
|
## 📋 BibTeX entry and citation info |
|
|
|
```bibtex |
|
Coming Soon |
|
``` |
