Upload config.py with huggingface_hub

config.py

@@ -0,0 +1,142 @@
+"""
+Path Configuration Module
+Handles all path resolution for the OmniParser system.
+Works on any machine regardless of installation location.
+"""
+
+import os
+from pathlib import Path
+import sys
+
+# Get the project root dynamically
+def get_project_root():
+    """
+    Get the project root directory.
+    Works by finding the parent of OmniParser directory.
+    """
+    # If this config is in project root, use that
+    config_file = Path(__file__).resolve()
+    
+    # Look for OmniParser directory to determine project root
+    current_dir = config_file.parent
+    
+    # Search up the directory tree for OmniParser or omoi-v2
+    for _ in range(5):  # Search up to 5 levels
+        if (current_dir / 'OmniParser').exists():
+            return current_dir
+        if (current_dir / '.git').exists() and 'omoi' in str(current_dir).lower():
+            return current_dir
+        current_dir = current_dir.parent
+    
+    # Fallback: assume omoi-v2 is in the parent of OmniParser
+    omniparser_dir = Path(__file__).parent.parent
+    if omniparser_dir.name == 'OmniParser':
+        return omniparser_dir.parent
+    
+    # Last resort: return current working directory
+    return Path.cwd()
+
+# Get all paths
+PROJECT_ROOT = get_project_root()
+OMNIPARSER_DIR = PROJECT_ROOT / 'OmniParser'
+WEIGHTS_DIR = OMNIPARSER_DIR / 'weights'
+ICON_DETECT_DIR = WEIGHTS_DIR / 'icon_detect'
+ICON_CAPTION_DIR = WEIGHTS_DIR / 'icon_caption_florence'
+
+# Data directories
+DATA_DIR = PROJECT_ROOT / 'data'
+CROPPED_IMAGES_DIR = PROJECT_ROOT / 'cropped_images'
+TEMP_CROP_DIR = Path('/tmp') / 'omoi_cropped_images'
+
+# Output directories
+OUTPUT_DIR = PROJECT_ROOT / 'output'
+
+# Create directories if they don't exist
+for directory in [DATA_DIR, CROPPED_IMAGES_DIR, OUTPUT_DIR, TEMP_CROP_DIR]:
+    directory.mkdir(parents=True, exist_ok=True)
+
+# ============ Model Paths ============
+
+def get_icon_detect_model_path():
+    """Get YOLOv8 icon detection model path"""
+    model_path = ICON_DETECT_DIR / 'model.pt'
+    if not model_path.exists():
+        raise FileNotFoundError(f"Icon detect model not found at: {model_path}")
+    return str(model_path)
+
+def get_icon_caption_model_path():
+    """Get Florence-2 caption model path (if it exists)"""
+    model_path = ICON_CAPTION_DIR / 'model.safetensors'
+    if model_path.exists():
+        return str(ICON_CAPTION_DIR)
+    return None
+
+# ============ Configuration ============
+
+def get_omniparser_config():
+    """Get default OmniParser configuration"""
+    return {
+        'som_model_path': get_icon_detect_model_path(),
+        'caption_model_name': None,  # Florence removed
+        'caption_model_path': get_icon_caption_model_path(),
+        'device': 'cpu',
+        'BOX_TRESHOLD': 0.05,
+        'save_cropped_images': True,
+        'cropped_images_dir': str(TEMP_CROP_DIR)
+    }
+
+# ============ Path Resolution Functions ============
+
+def resolve_path(path_str, base_dir=None):
+    """
+    Resolve a path that could be relative or absolute.
+    
+    Args:
+        path_str: Path string (relative or absolute)
+        base_dir: Base directory for relative paths (defaults to PROJECT_ROOT)
+    
+    Returns:
+        Absolute path as string
+    """
+    if base_dir is None:
+        base_dir = PROJECT_ROOT
+    
+    path = Path(path_str)
+    
+    # If already absolute, return as is
+    if path.is_absolute():
+        return str(path)
+    
+    # Relative path: resolve from base directory
+    resolved = (Path(base_dir) / path).resolve()
+    return str(resolved)
+
+def get_screenshot_path(filename='Screenshot2.png'):
+    """Get path to a screenshot file"""
+    return str(DATA_DIR / filename)
+
+def get_output_path(filename):
+    """Get path to an output file"""
+    return str(OUTPUT_DIR / filename)
+
+# ============ Print Configuration for Debugging ============
+
+if __name__ == "__main__":
+    print("\n" + "="*70)
+    print("OmniParser Configuration")
+    print("="*70)
+    print(f"\nProject Root: {PROJECT_ROOT}")
+    print(f"OmniParser Dir: {OMNIPARSER_DIR}")
+    print(f"Weights Dir: {WEIGHTS_DIR}")
+    print(f"\n[Model Paths]")
+    print(f"  Icon Detect: {get_icon_detect_model_path()}")
+    print(f"  Caption Model: {get_icon_caption_model_path()}")
+    print(f"\n[Data Directories]")
+    print(f"  Cropped Images: {CROPPED_IMAGES_DIR}")
+    print(f"  Temp Crop: {TEMP_CROP_DIR}")
+    print(f"  Output: {OUTPUT_DIR}")
+    print(f"\n[Configuration]")
+    config = get_omniparser_config()
+    for key, value in config.items():
+        print(f"  {key}: {value}")
+    print("\n" + "="*70 + "\n")