app.py

from transformers import pipeline
from PIL import Image
import gradio as gr
import numpy as np

# Load the Hugging Face depth estimation pipelines
pipe_base = pipeline(task="depth-estimation", model="LiheYoung/depth-anything-base-hf")
pipe_small = pipeline(task="depth-estimation", model="LiheYoung/depth-anything-small-hf")
pipe_intel = pipeline(task="depth-estimation", model="Intel/dpt-swinv2-tiny-256")
pipe_beit = pipeline(task="depth-estimation", model="Intel/dpt-beit-base-384")

def estimate_depths(image):
    # Perform depth estimation with each pipeline
    depth_base = pipe_base(image)["depth"]
    depth_small = pipe_small(image)["depth"]
    depth_intel = pipe_intel(image)["depth"]
    depth_beit = pipe_beit(image)["depth"]
    
    # Normalize depths for visualization
    depth_base = normalize_depth(depth_base)
    depth_small = normalize_depth(depth_small)
    depth_intel = normalize_depth(depth_intel)
    depth_beit = normalize_depth(depth_beit)
    
    return depth_base, depth_small, depth_intel, depth_beit

def normalize_depth(depth_map):
    # Normalize depth map values to range [0, 255] for visualization
    normalized_depth = ((depth_map - depth_map.min()) / (depth_map.max() - depth_map.min())) * 255
    return normalized_depth.astype(np.uint8)

# Create a Gradio interface using Blocks
with gr.Blocks() as iface:
    gr.Markdown("# Multi-Model Depth Estimation\nUpload an image to get depth estimation maps from multiple models.")
    
    with gr.Row():
        input_image = gr.Image(type="pil", label="Input Image")
    
    with gr.Row():
        with gr.Column():
            output_base = gr.Image(type="numpy", label="LiheYoung/depth-anything-base-hf", interactive=False)
            output_small = gr.Image(type="numpy", label="LiheYoung/depth-anything-small-hf", interactive=False)
        with gr.Column():
            output_intel = gr.Image(type="numpy", label="Intel/dpt-swinv2-tiny-256", interactive=False)
            output_beit = gr.Image(type="numpy", label="Intel/dpt-beit-base-384", interactive=False)

    input_image.change(fn=estimate_depths, inputs=input_image, outputs=[output_base, output_small, output_intel, output_beit])

# Launch the Gradio app
iface.launch()