Add Docker support and Gradio web UI

Add containerized deployment with NVIDIA CUDA support and a Gradio-based
web interface for easier usage. Based on community PR apple#13.

Author: claude

.dockerignore

@@ -0,0 +1,6 @@
+.dockerignore
+.gitignore
+*.md
+Dockerfile
+compose.yml
+data/

Dockerfile

@@ -0,0 +1,30 @@
+FROM nvidia/cuda:12.9.1-cudnn-devel-ubuntu24.04
+
+# Install Python 3.13
+RUN apt-get update && apt-get install -y wget software-properties-common build-essential && apt-get clean && rm -rf /var/lib/apt/lists/*
+RUN add-apt-repository ppa:deadsnakes/ppa
+RUN apt-get update && apt-get install -y python3.13 python3.13-venv python3.13-dev ninja-build && apt-get clean && rm -rf /var/lib/apt/lists/*
+
+# Install Sharp and dependencies
+RUN mkdir /app
+COPY pyproject.toml requirements.txt requirements.in /app/
+COPY src/ /app/src/
+WORKDIR /app
+RUN python3.13 -m venv .venv
+ENV TORCH_CUDA_ARCH_LIST="8.0;8.6;8.7;8.9;9.0+PTX"
+ENV FORCE_CUDA="1"
+RUN .venv/bin/pip install ninja
+RUN .venv/bin/pip install -r requirements.txt
+RUN .venv/bin/pip install gradio
+RUN ln -s /app/.venv/bin/sharp /usr/local/bin/sharp
+
+# Test run to download model and check if it works
+RUN wget https://apple.github.io/ml-sharp/thumbnails/Unsplash_-5wkyNA2BPc_0000-0001.jpg -O /tmp/test.jpg
+RUN sharp predict -i /tmp/test.jpg -o /tmp/test
+RUN rm /tmp/test.jpg /tmp/test -rf
+
+# Copy other files
+COPY . /app
+
+# Start Gradio web server
+CMD [".venv/bin/python3.13", "-u", "/app/gradio_web.py"]

compose.yml

@@ -0,0 +1,16 @@
+services:
+  sharp:
+    build:
+      context: .
+      dockerfile: Dockerfile
+    volumes:
+      - ./data:/app/data
+    ports:
+      - "7860:7860"
+    deploy:
+      resources:
+        reservations:
+          devices:
+            - driver: nvidia
+              count: 1
+              capabilities: [gpu]

gradio_web.py

@@ -0,0 +1,66 @@
+import gradio as gr
+import subprocess
+import os
+import shutil
+import time
+
+
+def predict(image):
+    # Ensure data directory exists
+    os.makedirs("/app/data", exist_ok=True)
+
+    input_path = "/app/data/input.jpg"
+
+    # Save/Copy input image
+    # image provided by gradio (type='filepath') is a temp path
+    shutil.copy(image, input_path)
+
+    # Run sharp command
+    # sharp predict -i /app/data/input.jpg -o /app/data/output --render
+    cmd = [
+        "sharp",
+        "predict",
+        "-i",
+        input_path,
+        "-o",
+        "/app/data/output",
+        "--render",
+    ]
+
+    # Execute command
+    try:
+        t = time.time()
+        print("Sharp started")
+        subprocess.run(cmd, check=True, capture_output=True)
+        print(f"Sharp command took {round(time.time() - t, 3)} seconds")
+    except subprocess.CalledProcessError as e:
+        print(f"Error running sharp: {e}")
+        print(f"Stdout: {e.stdout.decode()}")
+        print(f"Stderr: {e.stderr.decode()}")
+        return None
+
+    # Find output videos
+    rgb_video = "/app/data/output/input.mp4"
+    depth_video = "/app/data/output/input.depth.mp4"
+
+    if os.path.exists(rgb_video) and os.path.exists(depth_video):
+        return rgb_video, depth_video
+    elif os.path.exists(rgb_video):
+        return rgb_video, None
+
+    return None, None
+
+
+demo = gr.Interface(
+    fn=predict,
+    inputs=gr.Image(type="filepath", label="Input Image"),
+    outputs=[gr.Video(label="RGB Video"), gr.Video(label="Depth Video")],
+    title="Sharp 3D View Synthesis",
+    description="Upload an image to generate a 3D view synthesis video.",
+)
+
+if __name__ == "__main__":
+    print(
+        "Sharp Monocular View Synthesis in Less Than a Second (https://github.com/apple/ml-sharp)"
+    )
+    demo.launch(server_name="0.0.0.0", server_port=7860)