blip-base-captioning-ft-hl-scenes

BLIP-base fine-tuned for Image Captioning on High-Level descriptions of Scenes

BLIP base trained on the HL dataset for scene generation of images

Model fine-tuning 🏋️‍

• Trained for 10 epochs
• lr: 5e−5
• Adam optimizer
• half-precision (fp16)

Test set metrics 🧾

| Cider  | SacreBLEU  | Rouge-L |
|--------|------------|---------|
| 116.70 |   26.46    |  35.30  |

Model in Action 🚀

import requests
from PIL import Image
from transformers import BlipProcessor, BlipForConditionalGeneration

processor = BlipProcessor.from_pretrained("michelecafagna26/blip-base-captioning-ft-hl-scenes")
model = BlipForConditionalGeneration.from_pretrained("michelecafagna26/blip-base-captioning-ft-hl-scenes").to("cuda")

img_url = 'https://datasets-server.huggingface.co/assets/michelecafagna26/hl/--/default/train/0/image/image.jpg' 
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')


inputs = processor(raw_image, return_tensors="pt").to("cuda")
pixel_values = inputs.pixel_values

generated_ids = model.generate(pixel_values=pixel_values, max_length=50,
            do_sample=True,
            top_k=120,
            top_p=0.9,
            early_stopping=True,
            num_return_sequences=1)

processor.batch_decode(generated_ids, skip_special_tokens=True)

>>> "the picture is taken near a lake."

BibTex and citation info

@inproceedings{cafagna2023hl,
  title={{HL} {D}ataset: {V}isually-grounded {D}escription of {S}cenes, {A}ctions and
{R}ationales},
  author={Cafagna, Michele and van Deemter, Kees and Gatt, Albert},
  booktitle={Proceedings of the 16th International Natural Language Generation Conference (INLG'23)},
address = {Prague, Czech Republic},
  year={2023}
}