Best Model ViTB16 GPT2

🚀 Image Captioning with Vision Transformer

This project focuses on image captioning using a Vision Transformer (ViT) model. It provides a sample code to generate captions for images, supporting multiple languages and using specific evaluation metrics.

🚀 Quick Start

Prerequisites

• Install necessary libraries:
  • transformers
  • torch
  • Pillow

Installation

pip install transformers torch Pillow

💻 Usage Examples

Basic Usage

from transformers import VisionEncoderDecoderModel, ViTImageProcessor, GPT2Tokenizer
import torch
from PIL import Image

model = VisionEncoderDecoderModel.from_pretrained("evlinzxxx/best_model_ViTB16_GPT2")
feature_extractor = ViTImageProcessor.from_pretrained("evlinzxxx/best_model_ViTB16_GPT2")
tokenizer = GPT2Tokenizer.from_pretrained("evlinzxxx/best_model_ViTB16_GPT2")

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)

def show_image_and_captions(url):
  # get the image and display it
  display(load_image(url))
  # get the captions on various models
  our_caption = get_caption(model, image_processor, tokenizer, url)
  # print the captions
  print(f"Our caption: {our_caption}")

show_image_and_captions("/content/drive/MyDrive/try/test_400/gl_16.jpg") # ['navigate around the obstacle ahead adjusting your route to bypass the parked car.']

📚 Documentation

Metrics

Tags

• image-to-text
• image-captioning
• vision-transformer
• ViT-B/16

Supported Languages

• Indonesian (id)
• English (en)