File size: 791 Bytes
fd2c1e0 c13ea66 105b214 c13ea66 02a2188 c13ea66 1e62ffa 02a2188 c13ea66 fd2c1e0 c13ea66 105b214 479108b fd2c1e0 2fc3055 c13ea66 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 |
import gradio as gr
import torch
import clip
from PIL import Image
device = "cuda" if torch.cuda.is_available() else "cpu"
model, preprocess = clip.load("ViT-B/32", device=device)
def process_image_and_text(image, text):
text_list = text.tolist()
image = preprocess(image).unsqueeze(0).to(device)
text_tokens = clip.tokenize(text_list).to(device)
with torch.no_grad():
image_features = model.encode_image(image)
text_features = model.encode_text(text_tokens)
logits_per_image, logits_per_text = model(image, text_tokens)
probs = logits_per_image.softmax(dim=-1)
return probs
demo = gr.Interface(fn=process_image_and_text, inputs=[gr.inputs.Image(type="pil"), gr.inputs.Textbox()], outputs="text")
demo.launch()
|