File size: 4,300 Bytes
e0a1fd2
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
import torch
import torch.nn.functional as F
import cv2
import numpy as np
from PIL import Image
from pathlib import Path
import asyncio
from concurrent.futures import ThreadPoolExecutor
import gc

class VideoProcessor:
    def __init__(self):
        # Use CPU if no GPU
        self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
        print(f"Using device: {self.device}")
        
        # Load MiDaS (small model for speed)
        self.model = torch.hub.load("intel-isl/MiDaS", "MiDaS_small")
        self.model.to(self.device)
        self.model.eval()
        
        # Load transforms
        midas_transforms = torch.hub.load("intel-isl/MiDaS", "transforms")
        self.transform = midas_transforms.small_transform
        
        self.executor = ThreadPoolExecutor(max_workers=1)
    
    def hex_to_rgb(self, hex_color: str):
        """Convert hex to RGB"""
        hex_color = hex_color.lstrip('#')
        return tuple(int(hex_color[i:i+2], 16) for i in (0, 2, 4))
    
    async def process_video(self, input_path: str, threshold: float, 

                           bg_color: str, session_id: str) -> str:
        """Process video asynchronously"""
        loop = asyncio.get_event_loop()
        output_path = str(Path("/tmp") / f"{session_id}_output.mp4")
        
        # Run in thread pool
        await loop.run_in_executor(
            self.executor,
            self._process_video_sync,
            input_path, output_path, threshold, bg_color
        )
        
        return output_path
    
    def _process_video_sync(self, input_path: str, output_path: str,

                            threshold: float, bg_color: str):
        """Synchronous video processing"""
        cap = cv2.VideoCapture(input_path)
        fps = int(cap.get(cv2.CAP_PROP_FPS))
        width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
        height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
        total_frames = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
        
        # Output video
        fourcc = cv2.VideoWriter_fourcc(*'mp4v')
        out = cv2.VideoWriter(output_path, fourcc, fps, (width, height))
        
        bg_rgb = self.hex_to_rgb(bg_color)
        frame_count = 0
        
        while cap.isOpened():
            ret, frame = cap.read()
            if not ret:
                break
            
            # Process frame
            processed = self.process_frame(frame, threshold, bg_rgb)
            out.write(processed)
            
            frame_count += 1
            if frame_count % 30 == 0:
                print(f"Progress: {frame_count}/{total_frames}")
            
            # Clear cache occasionally
            if frame_count % 100 == 0:
                gc.collect()
                if torch.cuda.is_available():
                    torch.cuda.empty_cache()
        
        cap.release()
        out.release()
    
    def process_frame(self, frame: np.ndarray, threshold: float, 

                      bg_color: tuple) -> np.ndarray:
        """Process a single frame"""
        # Resize for speed
        h, w = frame.shape[:2]
        new_h, new_w = 256, int(256 * w / h)
        
        frame_small = cv2.resize(frame, (new_w, new_h))
        frame_rgb = cv2.cvtColor(frame_small, cv2.COLOR_BGR2RGB)
        
        # Get depth map
        img = Image.fromarray(frame_rgb)
        input_batch = self.transform(img).to(self.device)
        
        with torch.no_grad():
            depth = self.model(input_batch)
            depth = F.interpolate(
                depth.unsqueeze(1),
                size=(new_h, new_w),
                mode="bicubic",
                align_corners=False
            ).squeeze().cpu().numpy()
        
        # Normalize depth
        depth_norm = (depth - depth.min()) / (depth.max() - depth.min() + 1e-8)
        
        # Create mask and resize to original
        mask = (depth_norm > threshold).astype(np.uint8) * 255
        mask = cv2.resize(mask, (w, h), interpolation=cv2.INTER_LINEAR)
        mask = mask.astype(bool)
        
        # Apply background
        result = frame.copy()
        result[~mask] = bg_color
        
        return result