Files
test2/vr180_streaming/main.py
2025-07-27 08:01:08 -07:00

298 lines
8.4 KiB
Python

#!/usr/bin/env python3
"""
VR180 Streaming Human Matting - Main CLI entry point
"""
import argparse
import sys
from pathlib import Path
import traceback
from .config import StreamingConfig
from .streaming_processor import VR180StreamingProcessor
def create_parser() -> argparse.ArgumentParser:
"""Create command line argument parser"""
parser = argparse.ArgumentParser(
description="VR180 Streaming Human Matting - True streaming implementation",
formatter_class=argparse.RawDescriptionHelpFormatter,
epilog="""
Examples:
# Process video with streaming
vr180-streaming config-streaming.yaml
# Process with custom output
vr180-streaming config-streaming.yaml --output /path/to/output.mp4
# Generate example config
vr180-streaming --generate-config config-streaming-example.yaml
# Process specific frame range
vr180-streaming config-streaming.yaml --start-frame 1000 --max-frames 5000
"""
)
parser.add_argument(
"config",
nargs="?",
help="Path to YAML configuration file"
)
parser.add_argument(
"--generate-config",
metavar="PATH",
help="Generate example configuration file at specified path"
)
parser.add_argument(
"--output", "-o",
metavar="PATH",
help="Override output path from config"
)
parser.add_argument(
"--scale",
type=float,
metavar="FACTOR",
help="Override scale factor (0.25, 0.5, 1.0)"
)
parser.add_argument(
"--start-frame",
type=int,
metavar="N",
help="Start processing from frame N"
)
parser.add_argument(
"--max-frames",
type=int,
metavar="N",
help="Process at most N frames"
)
parser.add_argument(
"--device",
choices=["cuda", "cpu"],
help="Override processing device"
)
parser.add_argument(
"--format",
choices=["alpha", "greenscreen"],
help="Override output format"
)
parser.add_argument(
"--no-audio",
action="store_true",
help="Don't copy audio to output"
)
parser.add_argument(
"--verbose", "-v",
action="store_true",
help="Enable verbose output"
)
parser.add_argument(
"--dry-run",
action="store_true",
help="Validate configuration without processing"
)
return parser
def generate_example_config(output_path: str) -> None:
"""Generate example configuration file"""
config_content = '''# VR180 Streaming Configuration
# For RunPod or similar cloud GPU environments
input:
video_path: "/workspace/input_video.mp4"
start_frame: 0 # Start from beginning (or resume from checkpoint)
max_frames: null # Process entire video (or set limit for testing)
streaming:
mode: true # Enable streaming mode
buffer_frames: 10 # Small lookahead buffer
write_interval: 1 # Write every frame immediately
processing:
scale_factor: 0.5 # Process at 50% resolution for 8K input
adaptive_scaling: true # Dynamically adjust based on GPU load
target_gpu_usage: 0.7 # Target 70% GPU utilization
min_scale: 0.25
max_scale: 1.0
detection:
confidence_threshold: 0.7
model: "yolov8n" # Fast model for streaming
device: "cuda"
matting:
sam2_model_cfg: "sam2.1_hiera_l" # Large model for quality
sam2_checkpoint: "segment-anything-2/checkpoints/sam2.1_hiera_large.pt"
memory_offload: true # Essential for streaming
fp16: true # Use half precision
continuous_correction: true # Refine tracking periodically
correction_interval: 300 # Every 300 frames
stereo:
mode: "master_slave" # Left eye leads, right follows
master_eye: "left"
disparity_correction: true # Adjust for stereo depth
consistency_threshold: 0.3
baseline: 65.0 # mm - typical eye separation
focal_length: 1000.0 # pixels - adjust based on camera
output:
path: "/workspace/output_video.mp4"
format: "greenscreen" # or "alpha"
background_color: [0, 255, 0] # Pure green
video_codec: "h264_nvenc" # GPU encoding
quality_preset: "p4" # Balance quality/speed
crf: 18 # High quality
maintain_sbs: true # Keep side-by-side format
hardware:
device: "cuda"
max_vram_gb: 40.0 # RunPod A6000 has 48GB
max_ram_gb: 48.0 # Container RAM limit
recovery:
enable_checkpoints: true
checkpoint_interval: 1000 # Every 1000 frames
auto_resume: true # Resume from checkpoint if found
checkpoint_dir: "./checkpoints"
performance:
profile_enabled: true
log_interval: 100 # Log every 100 frames
memory_monitor: true # Track memory usage
'''
output_path = Path(output_path)
output_path.parent.mkdir(parents=True, exist_ok=True)
with open(output_path, 'w') as f:
f.write(config_content)
print(f"✅ Generated example configuration: {output_path}")
print("\nEdit the configuration file with your paths and run:")
print(f" python -m vr180_streaming {output_path}")
def validate_config(config: StreamingConfig, verbose: bool = False) -> bool:
"""Validate configuration and print any errors"""
errors = config.validate()
if errors:
print("❌ Configuration validation failed:")
for error in errors:
print(f" - {error}")
return False
if verbose:
print("✅ Configuration validation passed")
print(f" Input: {config.input.video_path}")
print(f" Output: {config.output.path}")
print(f" Scale: {config.processing.scale_factor}")
print(f" Device: {config.hardware.device}")
print(f" Format: {config.output.format}")
return True
def apply_cli_overrides(config: StreamingConfig, args: argparse.Namespace) -> None:
"""Apply command line overrides to configuration"""
if args.output:
config.output.path = args.output
if args.scale:
if not 0.1 <= args.scale <= 1.0:
raise ValueError("Scale factor must be between 0.1 and 1.0")
config.processing.scale_factor = args.scale
if args.start_frame is not None:
if args.start_frame < 0:
raise ValueError("Start frame must be non-negative")
config.input.start_frame = args.start_frame
if args.max_frames is not None:
if args.max_frames <= 0:
raise ValueError("Max frames must be positive")
config.input.max_frames = args.max_frames
if args.device:
config.hardware.device = args.device
config.detection.device = args.device
if args.format:
config.output.format = args.format
if args.no_audio:
config.output.maintain_sbs = False # This will skip audio copy
def main() -> int:
"""Main entry point"""
parser = create_parser()
args = parser.parse_args()
try:
# Handle config generation
if args.generate_config:
generate_example_config(args.generate_config)
return 0
# Require config file for processing
if not args.config:
parser.print_help()
print("\n❌ Error: Configuration file required")
print("\nGenerate an example config with:")
print(" vr180-streaming --generate-config config-streaming.yaml")
return 1
# Load configuration
config_path = Path(args.config)
if not config_path.exists():
print(f"❌ Error: Configuration file not found: {config_path}")
return 1
print(f"📄 Loading configuration from {config_path}")
config = StreamingConfig.from_yaml(str(config_path))
# Apply CLI overrides
apply_cli_overrides(config, args)
# Validate configuration
if not validate_config(config, verbose=args.verbose):
return 1
# Dry run mode
if args.dry_run:
print("✅ Dry run completed successfully")
return 0
# Process video
processor = VR180StreamingProcessor(config)
processor.process_video()
return 0
except KeyboardInterrupt:
print("\n⚠️ Processing interrupted by user")
return 130
except Exception as e:
print(f"\n❌ Error: {e}")
if args.verbose:
traceback.print_exc()
return 1
if __name__ == "__main__":
sys.exit(main())