Spaces:
Running
on
Zero
Running
on
Zero
| #!/usr/bin/env python | |
| from __future__ import annotations | |
| import gradio as gr | |
| import PIL.Image | |
| import spaces | |
| import torch | |
| from transformers import AutoProcessor, BlipForConditionalGeneration | |
| DESCRIPTION = "# Image Captioning with BLIP" | |
| device = torch.device("cuda" if torch.cuda.is_available() else "cpu") | |
| model_id = "Salesforce/blip-image-captioning-large" | |
| processor = AutoProcessor.from_pretrained(model_id) | |
| model = BlipForConditionalGeneration.from_pretrained(model_id).to(device) | |
| def run(image: PIL.Image.Image, text: str = "A picture of") -> str: | |
| inputs = processor(images=image, text=text, return_tensors="pt").to(device) | |
| generated_ids = model.generate(pixel_values=inputs.pixel_values, num_beams=3, max_length=20, min_length=5) | |
| generated_caption = processor.batch_decode(generated_ids, skip_special_tokens=True)[0] | |
| return generated_caption | |
| with gr.Blocks(css="style.css") as demo: | |
| gr.Markdown(DESCRIPTION) | |
| input_image = gr.Image(type="pil") | |
| text = gr.Textbox(label="Text", value="A picture of") | |
| run_button = gr.Button("Caption") | |
| output = gr.Textbox(label="Result") | |
| gr.on( | |
| triggers=[text.submit, run_button.click], | |
| fn=run, | |
| inputs=[input_image, text], | |
| outputs=output, | |
| api_name="caption", | |
| ) | |
| if __name__ == "__main__": | |
| demo.queue(max_size=20).launch() | |