|
from PIL import Image |
|
from transformers import BlipForConditionalGeneration, BlipProcessor |
|
|
|
processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-large") |
|
model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-large") |
|
|
|
def extract_image_details(image): |
|
inputs = processor(images=image, return_tensors="pt") |
|
|
|
generated_ids = model.generate( |
|
pixel_values=inputs["pixel_values"], |
|
max_length=50, |
|
num_beams=5, |
|
do_sample=False |
|
) |
|
generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0] |
|
print(f"BLIP Model Description: {generated_text}") |
|
return generated_text |
|
|
