Update app.py

app.py

@@ -1,58 +1,45 @@
 import gradio as gr
 from transformers import AutoProcessor, AutoModelForCausalLM
-import spaces
 import re
 from PIL import Image 
-
-import subprocess
-subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
+import os
+import numpy as np
+import space
 
 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:
-    """
-    Removes specific prefixes from captions if present, otherwise returns the original caption.
-    Args:
-        caption (str): A string containing a caption.
-    Returns:
-        str: The caption with the prefix removed if it was present, or the original caption.
-    """
-    # Define the prefixes to remove
-    prefix_substrings = [
-        ('captured from ', ''),
-        ('captured at ', '')
+    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"
     ]
     
-    # Create a regex pattern to match any of the prefixes
-    pattern = '|'.join([re.escape(opening) for opening, _ in prefix_substrings])
-    replacers = {opening.lower(): replacer for opening, replacer in prefix_substrings}
-    
-    # Function to replace matched prefix with its corresponding replacement
-    def replace_fn(match):
-        return replacers[match.group(0).lower()]
+    # 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)
     
-    # Apply the regex to the caption
-    modified_caption = re.sub(pattern, replace_fn, caption, count=1, 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())
     
-    # If the caption was modified, return the modified version; otherwise, return the original
-    return modified_caption if modified_caption != caption else caption
-    
-@spaces.GPU
-def run_example(image):
-    image = Image.fromarray(image)
-    task_prompt = "<MORE_DETAILED_CAPTION>"
-    prompt = task_prompt
+    return merged_content if merged_content != caption else caption
 
-    # Ensure the image is in RGB mode
+@space.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"],
@@ -61,37 +48,106 @@ def run_example(image):
         num_beams=3
     )
     generated_text = processor.batch_decode(generated_ids, skip_special_tokens=False)[0]
-    parsed_answer = processor.post_process_generation(generated_text, task=task_prompt, image_size=(image.width, image.height))
+    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; 
-  }
+#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="Florence-2 SD3 Prompts"):
+    
+    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=run_example,
+            inputs=[input_img],
+            outputs=[output_text],
+            fn=process_image,
             label='Try captioning on below examples'
-            )
+        )
+        
+        submit_btn.click(process_image, [input_img], [output_text])
 
-        submit_btn.click(run_example, [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)