pytorch-clip-interrogator 2023.2.19.10

Prompt engineering tool using BLIP 1/2 + CLIP Interrogate approach.

pytorch_clip_interrogator: Image-To-Promt.

Install package

pip install pytorch_clip_interrogator

Install the latest version

pip install --upgrade git+https://github.com/bes-dev/pytorch_clip_interrogator.git

Features

• Fully compatible with models from Huggingface.
• Supports BLIP 1/2 model.
• Support batch processing.

Usage

Simple code

import torch
import requests
from PIL import Image
from pytorch_clip_interrogator import PromptEngineer

# build pipeline
pipe = PromptEngineer(
    blip_model="Salesforce/blip2-opt-2.7b",
    clip_model="openai/clip-vit-base-patch32",
    device="cuda",
    torch_dtype=torch.float16
)

# load image
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')


# generate caption
print(pipe(image))