blip-base-captioning-ft-hl-narratives Open Source Model - Generate High-Level Narrative Image Descriptions for Free

Home

Blip Base Captioning Ft Hl Narratives

Developed by michelecafagna26

BLIP model fine-tuned on HL Narratives dataset for generating high-level narrative image descriptions

Image-to-Text

Transformers

EnglishOpen Source License:Apache-2.0 #Narrative image captioning #High-level semantic generation #Scene storytelling

Downloads 61

Release Time : 7/24/2023

Model Overview

This model is based on the BLIP architecture and fine-tuned on the HL Narratives dataset, specifically designed to generate rich narrative high-level descriptions from images rather than simple object recognition.

Model Features

Narrative description generation

Capable of generating high-level, narrative-rich image descriptions rather than simple object recognition

High-quality fine-tuning

Fine-tuned on specialized human narrative dataset (HL Narratives) to improve description quality

Multi-metric optimization

Performs well on multiple evaluation metrics including Cider, SacreBLEU and Rouge-L

Model Capabilities

Image caption generation

Visual language understanding

Narrative text generation

Use Cases

Content creation

Automatic image captioning

Automatically generates narrative-rich descriptions for image libraries

Produces more human-like and story-driven image captions

Assistive technology

Visual assistance

Provides detailed image descriptions for visually impaired users

Offers richer scene understanding than traditional image descriptions

🚀 BLIP-base fine-tuned for Narrative Image Captioning

This project presents a fine-tuned BLIP base model on the HL Narratives dataset, aiming to generate high-level narrative descriptions for images.

✨ Features

High-level Narrative Generation: Capable of generating high-level narrative descriptions for images.
Trained on Specific Dataset: Trained on the HL Narratives dataset.

🔧 Technical Details

Model fine-tuning 🏋️‍

Epochs: Trained for 3 epochs.
Learning Rate: 5e−5.
Optimizer: Adam optimizer.
Precision: Half-precision (fp16).

Test set metrics 🧾

Cider	SacreBLEU	Rouge-L
79.39	11.70	26.17

💻 Usage Examples

Basic Usage

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

processor = BlipProcessor.from_pretrained("blip-base-captioning-ft-hl-narratives")
model = BlipForConditionalGeneration.from_pretrained("blip-base-captioning-ft-hl-narratives").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)

>>> "she is holding an umbrella near a lake and is on vacation."

📄 License

This project is licensed under the Apache-2.0 license.

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}
}

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご