app.py

import os
# Set cache directories to writable locations right at the beginning
os.environ["HF_HOME"] = "/tmp/huggingface"
os.environ["TRANSFORMERS_CACHE"] = "/tmp/huggingface/transformers"
os.environ["MPLCONFIGDIR"] = "/tmp/matplotlib"

# Patching the schema handling problem in Gradio 5.x
# This needs to be done before any Gradio imports
import sys
def patch_gradio_utils():
    try:
        from gradio_client import utils
        original_get_type = utils.get_type
        
        def patched_get_type(schema):
            if isinstance(schema, bool):
                return "boolean"
            if not isinstance(schema, dict):
                return "any"
            return original_get_type(schema)
        
        utils.get_type = patched_get_type
        print("Successfully patched Gradio utils.get_type")
    except Exception as e:
        print(f"Could not patch Gradio utils: {e}")

patch_gradio_utils()

import gc
import torch
import cv2
import gradio as gr
print("📦 Gradio version:", gr.__version__)
import numpy as np
import matplotlib.cm as cm
import matplotlib  # New import for the updated colormap API
import subprocess

from video_depth_anything.video_depth import VideoDepthAnything
from utils.dc_utils import read_video_frames, save_video
from huggingface_hub import hf_hub_download

# Use GPU if available; otherwise, use CPU.
DEVICE = 'cuda' if torch.cuda.is_available() else 'cpu'

# Model configuration for different encoder variants.
model_configs = {
    'vits': {'encoder': 'vits', 'features': 64, 'out_channels': [48, 96, 192, 384]},
    'vitl': {'encoder': 'vitl', 'features': 256, 'out_channels': [256, 512, 1024, 1024]},
}
encoder2name = {
    'vits': 'Small',
    'vitl': 'Large',
}
encoder = 'vitl'
model_name = encoder2name[encoder]

# Initialize the model.
video_depth_anything = VideoDepthAnything(**model_configs[encoder])

filepath = hf_hub_download(
    repo_id=f"depth-anything/Video-Depth-Anything-{model_name}",
    filename=f"video_depth_anything_{encoder}.pth",
    repo_type="model",
    cache_dir="/tmp/huggingface"  # Explizites Setzen des Cache-Verzeichnisses
)
video_depth_anything.load_state_dict(torch.load(filepath, map_location='cpu'))
video_depth_anything = video_depth_anything.to(DEVICE).eval()

title = "# Video Depth Anything + RGBD sbs output"
description = """Official demo for **Video Depth Anything** + RGBD sbs output for viewing with Looking Glass Factory displays.
Please refer to our [paper](https://arxiv.org/abs/2501.12375), [project page](https://videodepthanything.github.io/), and [github](https://github.com/DepthAnything/Video-Depth-Anything) for more details."""

def infer_video_depth(
    input_video: str,
    max_len: int = -1,
    target_fps: int = -1,
    max_res: int = 1280,
    stitch: bool = True,
    grayscale: bool = True,
    convert_from_color: bool = True,
    blur: float = 0.3,
    loop_factor: int = 1,  # Neuer Parameter
    output_dir: str = './outputs',
    input_size: int = 518,
):
    # 1. Read input video frames for inference (downscaled to max_res).
    frames, target_fps = read_video_frames(input_video, max_len, target_fps, max_res)
    # 2. Perform depth inference using the model.
    depths, fps = video_depth_anything.infer_video_depth(frames, target_fps, input_size=input_size, device=DEVICE)

    video_name = os.path.basename(input_video)
    if not os.path.exists(output_dir):
        os.makedirs(output_dir)

    # Save the preprocessed (RGB) video and the generated depth visualization.
    # Still process the video, but we won't display it in the UI
    processed_video_path = os.path.join(output_dir, os.path.splitext(video_name)[0] + '_src.mp4')
    depth_vis_path = os.path.join(output_dir, os.path.splitext(video_name)[0] + '_vis.mp4')
    save_video(frames, processed_video_path, fps=fps)
    save_video(depths, depth_vis_path, fps=fps, is_depths=True)

    stitched_video_path = None
    if stitch:
        # For stitching: read the original video in full resolution (without downscaling).
        full_frames, _ = read_video_frames(input_video, max_len, target_fps, max_res=-1)
        # For each frame, create a visual depth image from the inferenced depths.
        d_min, d_max = depths.min(), depths.max()
        stitched_frames = []
        for i in range(min(len(full_frames), len(depths))):
            rgb_full = full_frames[i]  # Full-resolution RGB frame.
            depth_frame = depths[i]
            # Normalize the depth frame to the range [0, 255].
            depth_norm = ((depth_frame - d_min) / (d_max - d_min) * 255).astype(np.uint8)
            # Generate depth visualization:
            if grayscale:
                if convert_from_color:
                    # First, generate a color depth image using the inferno colormap,
                    # then convert that color image to grayscale.
                    cmap = matplotlib.colormaps.get_cmap("inferno")
                    depth_color = (cmap(depth_norm / 255.0)[..., :3] * 255).astype(np.uint8)
                    depth_gray = cv2.cvtColor(depth_color, cv2.COLOR_RGB2GRAY)
                    depth_vis = np.stack([depth_gray] * 3, axis=-1)
                else:
                    # Directly generate a grayscale image from the normalized depth values.
                    depth_vis = np.stack([depth_norm] * 3, axis=-1)
            else:
                # Generate a color depth image using the inferno colormap.
                cmap = matplotlib.colormaps.get_cmap("inferno")
                depth_vis = (cmap(depth_norm / 255.0)[..., :3] * 255).astype(np.uint8)
            # Apply Gaussian blur if requested.
            if blur > 0:
                kernel_size = int(blur * 20) * 2 + 1  # Ensures an odd kernel size.
                depth_vis = cv2.GaussianBlur(depth_vis, (kernel_size, kernel_size), 0)
            # Resize the depth visualization to match the full-resolution RGB frame.
            H_full, W_full = rgb_full.shape[:2]
            depth_vis_resized = cv2.resize(depth_vis, (W_full, H_full))
            # Concatenate the full-resolution RGB frame (left) and the resized depth visualization (right).
            stitched = cv2.hconcat([rgb_full, depth_vis_resized])
            stitched_frames.append(stitched)
        stitched_frames = np.array(stitched_frames)
        # Use only the first 20 characters of the base name for the output filename and append '_RGBD.mp4'
        base_name = os.path.splitext(video_name)[0]
        short_name = base_name[:20]
        stitched_video_path = os.path.join(output_dir, short_name + '_RGBD.mp4')
        save_video(stitched_frames, stitched_video_path, fps=fps)
        
        # Merge audio from the input video into the stitched video using ffmpeg.
        temp_audio_path = stitched_video_path.replace('_RGBD.mp4', '_RGBD_audio.mp4')
        cmd = [
            "ffmpeg",
            "-y",
            "-i", stitched_video_path,
            "-i", input_video,
            "-c:v", "copy",
            "-c:a", "aac",
            "-map", "0:v:0",
            "-map", "1:a:0?",
            "-shortest",
            temp_audio_path
        ]
        subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
        os.replace(temp_audio_path, stitched_video_path)
    
    # Nachdem die Videos erstellt wurden, wenden wir den Loop-Faktor an
    if loop_factor > 1:
        depth_looped_path = os.path.join(output_dir, os.path.splitext(os.path.basename(depth_vis_path))[0] + f'_loop{loop_factor}.mp4')
        
        # Erstelle eine temporäre Textdatei mit der Liste der zu wiederholenden Dateien
        concat_file_path = os.path.join(output_dir, 'concat_list.txt')
        with open(concat_file_path, 'w') as f:
            for _ in range(loop_factor):
                f.write(f"file '{depth_vis_path}'\n")
        
        # Verwende ffmpeg, um das Video zu wiederholen ohne Neucodierung
        cmd = [
            "ffmpeg",
            "-y",
            "-f", "concat",
            "-safe", "0",
            "-i", concat_file_path,
            "-c", "copy",
            depth_looped_path
        ]
        subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
        
        # Ersetze den ursprünglichen Pfad durch den neuen geloopten Pfad
        depth_vis_path = depth_looped_path
        
        if stitch and stitched_video_path:
            # Speichern wir den Originalnamen
            original_path = stitched_video_path
            
            # Überprüfen wir, ob das Input-Video einen Audio-Stream hat
            has_audio = False
            check_audio_cmd = [
                "ffmpeg",
                "-i", input_video,
                "-c", "copy",
                "-f", "null",
                "-"
            ]
            result = subprocess.run(check_audio_cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
            stderr = result.stderr.decode('utf-8')
            if "Audio" in stderr:
                has_audio = True
            
            # Erstelle eine temporäre Textdatei für die stitched Videos
            concat_stitched_file_path = os.path.join(output_dir, 'concat_stitched_list.txt')
            with open(concat_stitched_file_path, 'w') as f:
                for _ in range(loop_factor):
                    f.write(f"file '{original_path}'\n")
            
            # Temporärer Pfad für das geloopte Video ohne Audio
            temp_looped_path = os.path.join(output_dir, 'temp_looped_rgbd.mp4')
            
            # Verwende ffmpeg, um das Video zu loopen
            concat_cmd = [
                "ffmpeg",
                "-y",
                "-f", "concat",
                "-safe", "0",
                "-i", concat_stitched_file_path,
                "-c", "copy",
                temp_looped_path
            ]
            subprocess.run(concat_cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
            
            # Wenn Audio vorhanden ist, müssen wir es separat behandeln
            if has_audio:
                # Extrahiere den Audio-Track aus dem originalen Input-Video
                # Dies ist die sauberste Quelle
                audio_path = os.path.join(output_dir, 'extracted_audio.aac')
                extract_audio_cmd = [
                    "ffmpeg",
                    "-y",
                    "-i", input_video,  # Original Input-Video verwenden
                    "-vn", "-acodec", "copy",
                    audio_path
                ]
                subprocess.run(extract_audio_cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
                
                # Erstelle eine Textdatei für das Audio-Looping
                concat_audio_file_path = os.path.join(output_dir, 'concat_audio_list.txt')
                with open(concat_audio_file_path, 'w') as f:
                    for _ in range(loop_factor):
                        f.write(f"file '{audio_path}'\n")
                
                # Erstelle den geloopten Audio-Track
                looped_audio_path = os.path.join(output_dir, 'looped_audio.aac')
                audio_loop_cmd = [
                    "ffmpeg",
                    "-y",
                    "-f", "concat",
                    "-safe", "0",
                    "-i", concat_audio_file_path,
                    "-c", "copy",
                    looped_audio_path
                ]
                subprocess.run(audio_loop_cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
                
                # Kombiniere das geloopte Video mit dem geloopten Audio
                final_cmd = [
                    "ffmpeg",
                    "-y",
                    "-i", temp_looped_path,
                    "-i", looped_audio_path,
                    "-c:v", "copy",
                    "-c:a", "aac",
                    "-map", "0:v:0",
                    "-map", "1:a:0",
                    "-shortest",
                    original_path  # Verwenden des originalen Pfads als Ziel
                ]
                subprocess.run(final_cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
            else:
                # Wenn kein Audio vorhanden ist, einfach das Video umbenennen
                os.replace(temp_looped_path, original_path)
            
            # Bereinige temporäre Dateien
            temp_files = [concat_file_path, concat_stitched_file_path]
            if has_audio:
                temp_files.extend([concat_audio_file_path, audio_path, looped_audio_path])
            if os.path.exists(temp_looped_path):
                temp_files.append(temp_looped_path)
                
            for file_path in temp_files:
                if os.path.exists(file_path):
                    try:
                        os.remove(file_path)
                    except:
                        pass

    gc.collect()
    torch.cuda.empty_cache()

    # Only return the depth visualization and stitched video (not the preprocessed video)
    return [depth_vis_path, stitched_video_path]

def construct_demo():
    with gr.Blocks(analytics_enabled=False) as demo:
        gr.Markdown(title)
        gr.Markdown(description)
        gr.Markdown("### If you find this work useful, please help ⭐ the [Github Repo](https://github.com/DepthAnything/Video-Depth-Anything). Thanks for your attention!")
        
        with gr.Row(equal_height=True):
            with gr.Column(scale=1):
                # Video input component for file upload.
                input_video = gr.Video(label="Input Video")
            with gr.Column(scale=2):
                with gr.Row(equal_height=True):
                    # Removed the processed_video component from the UI
                    depth_vis_video = gr.Video(label="Generated Depth Video", interactive=False, autoplay=True, show_share_button=True, scale=5)
                    stitched_video = gr.Video(label="Stitched RGBD Video", interactive=False, autoplay=True, show_share_button=True, scale=5)
                    
        with gr.Row(equal_height=True):
            with gr.Column(scale=1):
                with gr.Accordion("Advanced Settings", open=False):
                    max_len = gr.Slider(label="Max process length", minimum=-1, maximum=1000, value=-1, step=1)
                    target_fps = gr.Slider(label="Target FPS", minimum=-1, maximum=30, value=-1, step=1)
                    max_res = gr.Slider(label="Max side resolution", minimum=480, maximum=1920, value=1280, step=1)
                    stitch_option = gr.Checkbox(label="Stitch RGB & Depth Videos", value=True)
                    grayscale_option = gr.Checkbox(label="Output Depth as Grayscale", value=True)
                    convert_from_color_option = gr.Checkbox(label="Convert Grayscale from Color", value=True)
                    blur_slider = gr.Slider(minimum=0, maximum=1, step=0.01, label="Depth Blur (can reduce edge artifacts on display)", value=0.3)
                    # Füge den Loop-Faktor Slider hinzu
                    loop_factor = gr.Slider(label="Loop Factor (repeats the output video)", minimum=1, maximum=20, value=1, step=1)
                generate_btn = gr.Button("Generate")
            with gr.Column(scale=2):
                pass
        
        # Removed Examples block to improve loading time
        
        generate_btn.click(
            fn=infer_video_depth,
            inputs=[input_video, max_len, target_fps, max_res, stitch_option, grayscale_option, convert_from_color_option, blur_slider, loop_factor],  # loop_factor hinzugefügt
            outputs=[depth_vis_video, stitched_video],
        )
    
    return demo

if __name__ == "__main__":
    demo = construct_demo()
    demo.queue()  # Enable asynchronous processing.
    demo.launch(share=True, show_api=False)