fathindifa/food-caption-blip2

Food Caption BLIP2

This is a fine-tuned version of the BLIP2 model for food image captioning.

Model Details

• Base model: BLIP2-OPT-2.7B
• Fine-tuned on food images
• Dataset size: 60 images
• Training epochs: 15
• Hardware used: CPU
• Final loss: 0.0001
• Training date: 2024-03-15

Usage

from transformers import Blip2Processor, Blip2ForConditionalGeneration
from PIL import Image

processor = Blip2Processor.from_pretrained("fathindifa/food-caption-blip2")
model = Blip2ForConditionalGeneration.from_pretrained("fathindifa/food-caption-blip2")

# Load and preprocess image
image = Image.open("food_image.jpg").convert('RGB')
inputs = processor(images=image, return_tensors="pt")

# Generate caption
outputs = model.generate(**inputs, max_new_tokens=32)
caption = processor.batch_decode(outputs, skip_special_tokens=True)[0]
print(caption)