| 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 |
|
|
