app.py

import gradio as gr
import torch
import os
import tempfile
import shutil
import time
import ffmpeg
import numpy as np
from PIL import Image
import moviepy.editor as mp
from infer import lotus  # Import the depth model inference function
import logging
import io
from multiprocessing import Pool, cpu_count

# Set up logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

# Set device to use the L40s GPU explicitly
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")

# Preprocess the video to adjust resolution and frame rate
def preprocess_video(video_path, target_fps=24, max_resolution=(1920, 1080)):
    """Preprocess the video to resize and adjust its frame rate."""
    video = mp.VideoFileClip(video_path)
    
    # Resize video if it's larger than the target resolution
    if video.size[0] > max_resolution[0] or video.size[1] > max_resolution[1]:
        video = video.resize(height=max_resolution[1])
    
    # Adjust FPS if target_fps is specified
    if target_fps > 0:
        video = video.set_fps(target_fps)
    
    return video

# Process a single frame through the depth model
def process_frame(args):
    """Process a single frame through the depth model and return depth map."""
    frame_index, frame_data, seed = args
    try:
        # Set seeds for reproducibility
        torch.manual_seed(seed)
        np.random.seed(seed)
        
        # Convert frame data to PIL Image
        image = Image.fromarray(frame_data).convert('RGB')
        
        # Save image to an in-memory file
        img_bytes = io.BytesIO()
        image.save(img_bytes, format='PNG')
        img_bytes.seek(0)  # Reset file pointer to the beginning
        
        # Process through the depth model
        _, output_d = lotus(img_bytes, 'depth', seed, device)
        
        # Convert depth output to numpy array
        depth_array = np.array(output_d)
        return (frame_index, depth_array)
            
    except Exception as e:
        logger.error(f"Error processing frame {frame_index}: {e}")
        return (frame_index, None)

# Process video frames and generate depth maps
def process_video(video_path, fps=0, seed=0, num_workers=4):
    """Process video frames in parallel and generate depth maps."""
    # Create a persistent temporary directory
    temp_dir = tempfile.mkdtemp()
    try:
        start_time = time.time()
        
        # Preprocess the video
        video = preprocess_video(video_path, target_fps=fps)
        
        # Use original video FPS if not specified
        if fps == 0:
            fps = video.fps
                
        frames = list(video.iter_frames(fps=video.fps))
        total_frames = len(frames)
        
        logger.info(f"Processing {total_frames} frames at {fps} FPS...")
        
        # Create directory for frame sequence and outputs
        frames_dir = os.path.join(temp_dir, "frames")
        os.makedirs(frames_dir, exist_ok=True)
        
        # Prepare arguments for multiprocessing
        args_list = [(i, frames[i], seed) for i in range(total_frames)]
        
        # Use multiprocessing Pool to process frames in parallel
        with Pool(processes=num_workers) as pool:
            results = []
            for result in pool.imap_unordered(process_frame, args_list):
                frame_index, depth_map = result
                if depth_map is not None:
                    # Save frame
                    frame_path = os.path.join(frames_dir, f"frame_{frame_index:06d}.png")
                    Image.fromarray(depth_map.squeeze()).save(frame_path)
                    
                    # Update preview every 10% progress
                    if (frame_index + 1) % max(1, total_frames // 10) == 0:
                        elapsed_time = time.time() - start_time
                        progress = ((frame_index + 1) / total_frames) * 100
                        yield depth_map, None, None, f"Processed {frame_index + 1}/{total_frames} frames... ({progress:.2f}%) Elapsed: {elapsed_time:.2f}s"
                else:
                    logger.error(f"Frame {frame_index} failed to process.")
        
        logger.info("Creating output files...")
            
        # Create ZIP of frame sequence
        zip_filename = f"depth_frames_{int(time.time())}.zip"
        zip_path = os.path.join(temp_dir, zip_filename)
        shutil.make_archive(zip_path[:-4], 'zip', frames_dir)
            
        # Create MP4 video
        video_filename = f"depth_video_{int(time.time())}.mp4"
        output_video_path = os.path.join(temp_dir, video_filename)
            
        try:
            # FFmpeg settings for high-quality MP4
            (
                ffmpeg
                .input(os.path.join(frames_dir, 'frame_%06d.png'), pattern_type='sequence', framerate=fps)
                .output(output_video_path, vcodec='libx264', pix_fmt='yuv420p', crf=17)
                .run(overwrite_output=True, quiet=True)
            )
            logger.info("MP4 video created successfully!")
                
        except ffmpeg.Error as e:
            logger.error(f"Error creating video: {e.stderr.decode() if e.stderr else str(e)}")
            output_video_path = None
        
        total_time = time.time() - start_time
        logger.info("Processing complete!")
        
        # Yield the file paths
        yield None, zip_path, output_video_path, f"Processing complete! Total time: {total_time:.2f} seconds"
            
    except Exception as e:
        logger.error(f"Error: {e}")
        yield None, None, None, f"Error processing video: {e}"
    # Do not delete temp_dir here; we need the files to persist
    # Cleanup can be handled elsewhere if necessary

# Wrapper function with error handling
def process_wrapper(video, fps=0, seed=0, num_workers=4):
    if video is None:
        raise gr.Error("Please upload a video.")
    try:
        outputs = []
        # Use video directly, since it's the file path
        for output in process_video(video, fps, seed, num_workers):
            outputs.append(output)
            yield output
        return outputs[-1]
    except Exception as e:
        raise gr.Error(f"Error processing video: {str(e)}")

# Custom CSS for styling (unchanged)
custom_css = """
    /* Your existing custom CSS */
"""

# Gradio Interface
with gr.Blocks(css=custom_css) as demo:
    gr.HTML('''
        <div class="title-container">
            <div id="title">Video Depth Estimation</div>
        </div>
    ''')
    
    with gr.Row():
        with gr.Column():
            video_input = gr.Video(label="Upload Video", interactive=True)
            fps_slider = gr.Slider(minimum=0, maximum=60, step=1, value=0, label="Output FPS (0 for original)")
            seed_slider = gr.Number(value=0, label="Seed")
            num_workers_slider = gr.Slider(minimum=1, maximum=cpu_count(), step=1, value=4, label="Number of Workers")
            btn = gr.Button("Process Video")
        
        with gr.Column():
            preview_image = gr.Image(label="Live Preview")
            output_frames_zip = gr.File(label="Download Frame Sequence (ZIP)")
            output_video = gr.File(label="Download Video (MP4)")
            time_textbox = gr.Textbox(label="Status", interactive=False)
            
    btn.click(
        fn=process_wrapper,
        inputs=[video_input, fps_slider, seed_slider, num_workers_slider],
        outputs=[preview_image, output_frames_zip, output_video, time_textbox]
    )

    demo.queue()

if __name__ == "__main__":
    demo.launch(debug=True)