app.py

import gradio as gr
from transformers import AutoProcessor, AutoModelForCausalLM
import re
from PIL import Image 
import os
import numpy as np
import spaces

import subprocess
subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)

model = AutoModelForCausalLM.from_pretrained('microsoft/Florence-2-large', trust_remote_code=True).to("cuda").eval()
processor = AutoProcessor.from_pretrained('microsoft/Florence-2-large', trust_remote_code=True)

TITLE = "# [Florence-2 SD3 Long Captioner](https://huggingface.co/gokaygokay/Florence-2-SD3-Captioner/)"
DESCRIPTION = "[Florence-2 Base](https://huggingface.co/microsoft/Florence-2-base-ft) fine-tuned on Long SD3 Prompt and Image pairs. Check above link for datasets that are used for fine-tuning."

def modify_caption(caption: str) -> str:
    special_patterns = [
        (r'The image shows ', ''),      # 匹配 "The image shows " 并替换为空字符串
        (r'The image is .*? of ', ''),  # 匹配 "The image is .*? of" 并替换为空字符串
        (r'of the .*? is', 'is')        # 匹配 "of the .*? is" 并替换为 "is"
    ]
    
    # 对每个特殊模式进行替换
    for pattern, replacement in special_patterns:
        caption = re.sub(pattern, replacement, caption, flags=re.IGNORECASE)
    
    no_blank_lines = re.sub(r'\n\s*\n', '\n', caption)
    # 合并内容
    merged_content = ' '.join(no_blank_lines.strip().splitlines())
    
    return merged_content if merged_content != caption else caption

@spaces.GPU
def process_image(image):
    if isinstance(image, np.ndarray):
        image = Image.fromarray(image)
    elif isinstance(image, str):
        image = Image.open(image)
    if image.mode != "RGB":
        image = image.convert("RGB")
    
    prompt = "<MORE_DETAILED_CAPTION>"
    
    inputs = processor(text=prompt, images=image, return_tensors="pt").to("cuda")
    generated_ids = model.generate(
        input_ids=inputs["input_ids"],
        pixel_values=inputs["pixel_values"],
        max_new_tokens=1024,
        num_beams=3
    )
    generated_text = processor.batch_decode(generated_ids, skip_special_tokens=False)[0]
    parsed_answer = processor.post_process_generation(generated_text, task=prompt, image_size=(image.width, image.height))
    return modify_caption(parsed_answer["<MORE_DETAILED_CAPTION>"])

def extract_frames(image_path, output_folder):
    with Image.open(image_path) as img:
        base_name = os.path.splitext(os.path.basename(image_path))[0]
        frame_paths = []
        
        try:
            for i in range(0, img.n_frames):
                img.seek(i)
                frame_path = os.path.join(output_folder, f"{base_name}_frame_{i:03d}.png")
                img.save(frame_path)
                frame_paths.append(frame_path)
        except EOFError:
            pass  # We've reached the end of the sequence
        
        return frame_paths

def process_folder(folder_path):
    if not os.path.isdir(folder_path):
        return "Invalid folder path."
    
    processed_files = []
    skipped_files = []
    for filename in os.listdir(folder_path):
        if filename.lower().endswith(('.png', '.jpg', '.jpeg', '.gif', '.bmp', '.webp', '.heic')):
            image_path = os.path.join(folder_path, filename)
            txt_filename = os.path.splitext(filename)[0] + '.txt'
            txt_path = os.path.join(folder_path, txt_filename)
            
            # Check if the corresponding text file already exists
            if os.path.exists(txt_path):
                skipped_files.append(f"Skipped {filename} (text file already exists)")
                continue
            
            # Check if the image has multiple frames
            with Image.open(image_path) as img:
                if getattr(img, "is_animated", False) and img.n_frames > 1:
                    # Extract frames
                    frames = extract_frames(image_path, folder_path)
                    for frame_path in frames:
                        frame_txt_filename = os.path.splitext(os.path.basename(frame_path))[0] + '.txt'
                        frame_txt_path = os.path.join(folder_path, frame_txt_filename)
                        
                        # Check if the corresponding text file for the frame already exists
                        if os.path.exists(frame_txt_path):
                            skipped_files.append(f"Skipped {os.path.basename(frame_path)} (text file already exists)")
                            continue
                        
                        caption = process_image(frame_path)
                        
                        with open(frame_txt_path, 'w', encoding='utf-8') as f:
                            f.write(caption)
                        
                        processed_files.append(f"Processed {os.path.basename(frame_path)} -> {frame_txt_filename}")
                else:
                    # Process single image
                    caption = process_image(image_path)
                    
                    with open(txt_path, 'w', encoding='utf-8') as f:
                        f.write(caption)
                    
                    processed_files.append(f"Processed {filename} -> {txt_filename}")
    
    result = "\n".join(processed_files + skipped_files)
    return result if result else "No image files found or all files were skipped in the specified folder."

css = """
#output { height: 500px; overflow: auto; border: 1px solid #ccc; }
"""

with gr.Blocks(css=css) as demo:
    gr.Markdown(TITLE)
    gr.Markdown(DESCRIPTION)
    
    with gr.Tab(label="Single Image Processing"):
        with gr.Row():
            with gr.Column():
                input_img = gr.Image(label="Input Picture")
                submit_btn = gr.Button(value="Submit")
            with gr.Column():
                output_text = gr.Textbox(label="Output Text")
        
        gr.Examples(
            [["image1.jpg"], ["image2.jpg"], ["image3.png"], ["image4.jpg"], ["image5.jpg"], ["image6.PNG"]],
            inputs=[input_img],
            outputs=[output_text],
            fn=process_image,
            label='Try captioning on below examples'
        )
        
        submit_btn.click(process_image, [input_img], [output_text])

    with gr.Tab(label="Batch Processing"):
        with gr.Row():
            folder_input = gr.Textbox(label="Input Folder Path")
            batch_submit_btn = gr.Button(value="Process Folder")
        batch_output = gr.Textbox(label="Batch Processing Results", lines=10)
        
        batch_submit_btn.click(process_folder, [folder_input], [batch_output])

demo.launch(debug=True)