VideoBackgroundReplacer / debug_positioning.py
MogensR's picture
Update debug_positioning.py
ae33908 verified
#!/usr/bin/env python3
"""
Debug Enhanced Positioning System for Video Background Replacement
Key debugging areas:
1. UI parameter flow validation
2. Coordinate system debugging
3. Alpha/video alignment verification
4. Canvas placement mathematics
5. Temporal synchronization checks
"""
import cv2
import numpy as np
import os
import time
import random
from pathlib import Path
from moviepy.editor import VideoFileClip
import logging
# Enhanced logging for debugging
logging.basicConfig(level=logging.DEBUG, format="%(asctime)s - %(levelname)s - %(message)s")
debug_logger = logging.getLogger("positioning_debug")
class PositioningDebugger:
"""Comprehensive debugging for positioning system"""
def __init__(self, debug_dir="debug_positioning"):
self.debug_dir = Path(debug_dir)
self.debug_dir.mkdir(exist_ok=True, parents=True)
self.frame_count = 0
def log_parameters(self, placement_dict):
"""Log all positioning parameters"""
debug_logger.info("=== POSITIONING PARAMETERS ===")
debug_logger.info(f"Raw placement dict: {placement_dict}")
px = float(placement_dict.get("x", 0.5))
py = float(placement_dict.get("y", 0.75))
ps = float(placement_dict.get("scale", 1.0))
feather_px = int(placement_dict.get("feather", 3))
debug_logger.info(f"Parsed - px: {px}, py: {py}, ps: {ps}, feather: {feather_px}")
debug_logger.info(f"Clamped - px: {max(0.0, min(1.0, px))}, py: {max(0.0, min(1.0, py))}")
debug_logger.info(f"Scale range - ps: {max(0.3, min(2.0, ps))}")
return px, py, ps, feather_px
def debug_coordinate_calculation(self, frame_dims, alpha_dims, placement):
"""Debug coordinate calculations step by step"""
hh, ww = frame_dims # Original video dimensions
alpha_h, alpha_w = alpha_dims # Alpha video dimensions
px, py, ps, feather_px = self.log_parameters(placement)
debug_logger.info("=== COORDINATE CALCULATIONS ===")
debug_logger.info(f"Original video dims: {ww}x{hh}")
debug_logger.info(f"Alpha video dims: {alpha_w}x{alpha_h}")
# Scale calculations
sw = max(1, int(ww * ps))
sh = max(1, int(hh * ps))
debug_logger.info(f"Scaled subject size: {sw}x{sh} (scale factor: {ps})")
# Center calculations
cx = int(px * ww)
cy = int(py * hh)
debug_logger.info(f"Target center: ({cx}, {cy})")
# Top-left corner calculations
x0 = int(cx - sw // 2)
y0 = int(cy - sh // 2)
debug_logger.info(f"Top-left corner: ({x0}, {y0})")
# Clipping calculations
xs0, ys0 = max(0, x0), max(0, y0)
xs1, ys1 = min(ww, x0 + sw), min(hh, y0 + sh)
debug_logger.info(f"Clipped bounds: ({xs0}, {ys0}) to ({xs1}, {ys1})")
if xs1 <= xs0 or ys1 <= ys0:
debug_logger.warning("INVALID BOUNDS - Subject will be clipped completely!")
# Source region calculations
src_x0 = xs0 - x0
src_y0 = ys0 - y0
src_x1 = src_x0 + (xs1 - xs0)
src_y1 = src_y0 + (ys1 - ys0)
debug_logger.info(f"Source region: ({src_x0}, {src_y0}) to ({src_x1}, {src_y1})")
return {
'scaled_size': (sw, sh),
'center': (cx, cy),
'top_left': (x0, y0),
'clipped_dest': (xs0, ys0, xs1, ys1),
'source_region': (src_x0, src_y0, src_x1, src_y1)
}
def save_debug_frame(self, frame, alpha, composite, frame_idx, coords):
"""Save debug visualization of frame processing"""
debug_path = self.debug_dir / f"frame_{frame_idx:04d}_debug.png"
# Create debug visualization
h, w = frame.shape[:2]
debug_viz = np.zeros((h * 2, w * 2, 3), dtype=np.uint8)
# Top-left: Original frame
debug_viz[0:h, 0:w] = cv2.cvtColor((frame * 255).astype(np.uint8), cv2.COLOR_RGB2BGR)
# Top-right: Alpha mask (as RGB)
alpha_viz = cv2.cvtColor((alpha * 255).astype(np.uint8), cv2.COLOR_GRAY2BGR)
debug_viz[0:h, w:w*2] = alpha_viz
# Bottom-left: Composite
debug_viz[h:h*2, 0:w] = cv2.cvtColor((composite * 255).astype(np.uint8), cv2.COLOR_RGB2BGR)
# Bottom-right: Coordinate overlay
coord_viz = np.zeros((h, w, 3), dtype=np.uint8)
# Draw placement indicators
cx, cy = coords['center']
sw, sh = coords['scaled_size']
x0, y0 = coords['top_left']
# Draw center cross
cv2.line(coord_viz, (cx-10, cy), (cx+10, cy), (0, 255, 0), 2)
cv2.line(coord_viz, (cx, cy-10), (cx, cy+10), (0, 255, 0), 2)
# Draw scaled subject bounds
cv2.rectangle(coord_viz, (x0, y0), (x0 + sw, y0 + sh), (255, 0, 0), 2)
# Draw frame bounds
cv2.rectangle(coord_viz, (0, 0), (w-1, h-1), (255, 255, 255), 1)
debug_viz[h:h*2, w:w*2] = coord_viz
# Add text annotations
font = cv2.FONT_HERSHEY_SIMPLEX
cv2.putText(debug_viz, "Original", (10, 30), font, 0.7, (255, 255, 255), 2)
cv2.putText(debug_viz, "Alpha", (w + 10, 30), font, 0.7, (255, 255, 255), 2)
cv2.putText(debug_viz, "Composite", (10, h + 30), font, 0.7, (255, 255, 255), 2)
cv2.putText(debug_viz, "Coordinates", (w + 10, h + 30), font, 0.7, (255, 255, 255), 2)
cv2.imwrite(str(debug_path), debug_viz)
debug_logger.info(f"Debug frame saved: {debug_path}")
def create_enhanced_composite_function(debugger=None):
"""Create composite function with enhanced debugging"""
def composite_frame_debug(get_frame, t, original_clip, alpha_clip, bg_rgb, placement, feather_px):
"""Enhanced composite function with comprehensive debugging"""
if debugger:
debugger.frame_count += 1
debug_logger.info(f"=== PROCESSING FRAME {debugger.frame_count} at t={t:.3f}s ===")
# Get original frame
frame = get_frame(t).astype(np.float32) / 255.0
hh, ww = frame.shape[:2]
# Get alpha with temporal bounds checking
alpha_duration = alpha_clip.duration or 0
if alpha_duration > 0:
alpha_t = min(t, max(0.0, alpha_duration - 0.01))
else:
alpha_t = 0.0
debug_logger.info(f"Alpha lookup: t={t:.3f}, alpha_t={alpha_t:.3f}, alpha_duration={alpha_duration:.3f}")
try:
a = alpha_clip.get_frame(alpha_t)
if a.ndim == 3:
a = a[:, :, 0] # Take first channel if RGB
a = a.astype(np.float32) / 255.0
debug_logger.info(f"Alpha frame shape: {a.shape}, range: [{a.min():.3f}, {a.max():.3f}]")
except Exception as e:
debug_logger.error(f"Alpha frame error: {e}")
return (bg_rgb * 255).astype(np.uint8)
# Extract placement parameters
px = max(0.0, min(1.0, float(placement.get("x", 0.5))))
py = max(0.0, min(1.0, float(placement.get("y", 0.75))))
ps = max(0.3, min(2.0, float(placement.get("scale", 1.0))))
# Debug coordinate calculations
coords = None
if debugger:
coords = debugger.debug_coordinate_calculation((hh, ww), a.shape, placement)
# Scale subject and alpha
sw = max(1, int(ww * ps))
sh = max(1, int(hh * ps))
debug_logger.info(f"Scaling: {ww}x{hh} -> {sw}x{sh} (factor: {ps})")
fg_scaled = cv2.resize(frame, (sw, sh), interpolation=cv2.INTER_LINEAR)
a_scaled = cv2.resize(a, (sw, sh), interpolation=cv2.INTER_LINEAR)
# Create canvas
fg_canvas = np.zeros_like(frame, dtype=np.float32)
a_canvas = np.zeros((hh, ww), dtype=np.float32)
# Calculate placement
cx = int(px * ww)
cy = int(py * hh)
x0 = int(cx - sw // 2)
y0 = int(cy - sh // 2)
# Bounds checking
xs0, ys0 = max(0, x0), max(0, y0)
xs1, ys1 = min(ww, x0 + sw), min(hh, y0 + sh)
if xs1 <= xs0 or ys1 <= ys0:
debug_logger.warning("Subject completely outside frame bounds!")
if debugger:
debugger.save_debug_frame(frame, a, bg_rgb, debugger.frame_count, coords or {})
return (bg_rgb * 255).astype(np.uint8)
# Place scaled subject
src_x0 = xs0 - x0
src_y0 = ys0 - y0
src_x1 = src_x0 + (xs1 - xs0)
src_y1 = src_y0 + (ys1 - ys0)
try:
fg_canvas[ys0:ys1, xs0:xs1, :] = fg_scaled[src_y0:src_y1, src_x0:src_x1, :]
a_canvas[ys0:ys1, xs0:xs1] = a_scaled[src_y0:src_y1, src_x0:src_x1]
except Exception as e:
debug_logger.error(f"Canvas placement error: {e}")
debug_logger.error(f"Canvas region: [{ys0}:{ys1}, {xs0}:{xs1}]")
debug_logger.error(f"Source region: [{src_y0}:{src_y1}, {src_x0}:{src_x1}]")
if debugger:
debugger.save_debug_frame(frame, a, bg_rgb, debugger.frame_count, coords or {})
return (bg_rgb * 255).astype(np.uint8)
# Apply feathering
if feather_px > 0:
k = (feather_px * 2 + 1)
a_canvas = cv2.GaussianBlur(a_canvas, (k, k), feather_px)
# Composite
a3 = a_canvas[:, :, None]
comp = a3 * fg_canvas + (1.0 - a3) * bg_rgb
result = np.clip(comp * 255, 0, 255).astype(np.uint8)
# Save debug frame periodically
if debugger and debugger.frame_count % 30 == 1: # Every ~1 second at 30fps
debugger.save_debug_frame(frame, a_canvas, comp, debugger.frame_count, coords or {})
return result
return composite_frame_debug
def validate_ui_parameter_flow():
"""Test function to validate UI parameters reach processing correctly"""
# Simulate UI parameter values
test_placements = [
{"x": 0.5, "y": 0.5, "scale": 1.0, "feather": 3}, # Center
{"x": 0.2, "y": 0.8, "scale": 0.7, "feather": 5}, # Bottom-left, smaller
{"x": 0.8, "y": 0.3, "scale": 1.5, "feather": 1}, # Top-right, larger
]
debugger = PositioningDebugger()
for i, placement in enumerate(test_placements):
print(f"\n=== TEST CASE {i+1} ===")
# Simulate frame dimensions
frame_dims = (720, 1280) # H, W
alpha_dims = (720, 1280)
coords = debugger.debug_coordinate_calculation(frame_dims, alpha_dims, placement)
print(f"Expected center: {coords['center']}")
print(f"Scaled size: {coords['scaled_size']}")
print(f"Placement bounds: {coords['clipped_dest']}")
def create_debug_enhanced_process_video_main():
"""Enhanced version of process_video_main with debugging"""
def process_video_main_debug(
video_path: str,
background_path: str = None,
trim_duration: float = None,
crf: int = 18,
preserve_audio_flag: bool = True,
placement: dict = None,
use_chunked_processing: bool = False,
enable_debug: bool = True,
progress=None,
):
"""
Enhanced process_video_main with comprehensive positioning debugging
"""
debugger = PositioningDebugger() if enable_debug else None
messages = []
try:
if debugger:
debug_logger.info("=== STARTING DEBUG SESSION ===")
debug_logger.info(f"Video: {video_path}")
debug_logger.info(f"Background: {background_path}")
debug_logger.info(f"Placement: {placement}")
# Validate placement parameters early
placement = placement or {}
if debugger:
debugger.log_parameters(placement)
# ... [Previous initialization code remains the same] ...
# Enhanced composite function with debugging
def composite_frame(get_frame, t):
return create_enhanced_composite_function(debugger)(
get_frame, t, original_clip, alpha_clip, bg_rgb, placement,
int(placement.get("feather", 3))
)
# ... [Rest of processing remains the same] ...
if debugger:
debug_logger.info("=== DEBUG SESSION COMPLETE ===")
debug_logger.info(f"Debug files saved to: {debugger.debug_dir}")
return result_path, "\n".join(messages)
except Exception as e:
if debugger:
debug_logger.error(f"Processing failed: {e}")
raise
return process_video_main_debug
# Quick fix suggestions for immediate testing
def quick_positioning_fixes():
"""
Quick fixes to test for common positioning issues
"""
fixes = {
"coordinate_system_flip": {
"description": "Test if Y coordinate should be flipped",
"change": "cy = int((1.0 - py) * hh) # Flip Y coordinate",
"original": "cy = int(py * hh)"
},
"anchor_point_adjustment": {
"description": "Test different anchor points for scaling",
"change": """
# Try bottom-center anchor instead of center-center
x0 = int(cx - sw // 2) # Keep X centered
y0 = int(cy - sh) # Anchor at bottom
""",
"original": """
x0 = int(cx - sw // 2)
y0 = int(cy - sh // 2)
"""
},
"alpha_scaling_sync": {
"description": "Ensure alpha and frame scaling are synchronized",
"change": """
# Force exact same dimensions
if a.shape != (hh, ww):
a = cv2.resize(a, (ww, hh), interpolation=cv2.INTER_LINEAR)
""",
"original": "# No explicit resize check"
},
"ui_parameter_validation": {
"description": "Add parameter validation and logging",
"change": """
px = float(placement.get("x", 0.5))
py = float(placement.get("y", 0.75))
ps = float(placement.get("scale", 1.0))
print(f"DEBUG: px={px}, py={py}, ps={ps}")
print(f"DEBUG: cx={px*ww}, cy={py*hh}")
""",
"original": "# No debug logging"
}
}
return fixes
if __name__ == "__main__":
print("=== Video Background Replacement - Positioning Debug ===")
print("\n1. Running UI parameter validation...")
validate_ui_parameter_flow()
print("\n2. Available quick fixes:")
fixes = quick_positioning_fixes()
for fix_name, fix_info in fixes.items():
print(f"\n{fix_name.upper()}:")
print(f" Description: {fix_info['description']}")
print(f" Change: {fix_info['change']}")
print("\n3. Debug system ready!")
print(" - Use PositioningDebugger class in your main pipeline")
print(" - Enable debug mode: enable_debug=True")
print(" - Check debug_positioning/ folder for output")