app.py

import os
from PIL import Image

import gradio as gr
import numpy as np
import torch
from transformers import AutoModelForDepthEstimation, DPTImageProcessor


processor = DPTImageProcessor.from_pretrained(
    "Intel/dpt-large")
model = AutoModelForDepthEstimation.from_pretrained("Intel/dpt-large")


def main(image, input_size=384):
    # prepare image for the model
    inputs = processor(images=image, return_tensors="pt", do_resize=True, size=(
        input_size, input_size), keep_aspect_ratio=True)
    print(type(inputs), inputs.data["pixel_values"].shape)

    # do inference
    with torch.no_grad():
        outputs = model(**inputs)
        predicted_depth = outputs.predicted_depth

    # interpolate to original size
    prediction = torch.nn.functional.interpolate(predicted_depth.unsqueeze(
        1), size=image.shape[:-1], mode="bicubic").squeeze()
    output = prediction.cpu().numpy().copy()
    formatted = (output * 255 / output.max()).astype("uint8")
    depth = Image.fromarray(formatted)
    return depth


title = "Demo: monocular depth estimation with DPT"
description = "This demo uses <a href='https://huggingface.co/Intel/dpt-large' target='_blank'>DPT</a> to estimate depth from monocular image."
examples = [[f"examples/{file}"]
            for file in os.listdir("examples") if file[0] != "."]

demo = gr.Interface(fn=main, inputs=[gr.Image(label="Input Image"), gr.Slider(128, 512, value=384, label="Input Size")], outputs="image",
                    title=title, description=description, examples=examples, cache_examples=True)
demo.launch(debug=True, share=True)