app.py

import os
import gradio as gr
import numpy as np
import glob
import warnings
import pandas as pd
import matplotlib.pyplot as plt

from utils import OrthogonalRegularizer
from huggingface_hub.keras_mixin import from_pretrained_keras

# load model
model = from_pretrained_keras(
    "keras-io/pointnet_segmentation", custom_objects={"OrthogonalRegularizer": OrthogonalRegularizer}
)

# Examples
samples = []
input_images = glob.glob("asset/source/*.csv")
examples = [[im] for im in input_images]
LABELS = ["wing", "body", "tail", "engine"]
COLORS = ["blue", "green", "red", "pink"]


def visualize_data(point_cloud, labels, output_path=None):
    df = pd.DataFrame(
        data={
            "x": point_cloud[:, 0],
            "y": point_cloud[:, 1],
            "z": point_cloud[:, 2],
            "label": labels,
        }
    )
    fig = plt.figure(figsize=(15, 10))
    ax = plt.axes(projection="3d")
    for index, label in enumerate(LABELS):
        c_df = df[df["label"] == label]
        try:
            ax.scatter(c_df["x"], c_df["y"], c_df["z"], label=label, alpha=0.5, c=COLORS[index])
        except IndexError:
            pass
    ax.legend()
    if output_path:
        os.makedirs(os.path.dirname(output_path), exist_ok=True)
        plt.savefig(output_path)


def inference(
    csv_file,
    output_path="asset/output",
    cpu=False,
):

    csv_path = csv_file.name
    im_name = csv_path.split("/")[-1].split(".")[0]

    if os.path.exists(csv_path):
        df = pd.read_csv(csv_path, index_col=None)
        inputs = df[["x", "y", "z"]].values
        y_test = df.iloc[:, 3:].values  # TODO: show ground truth image if y_test is not None
    else:
        warnings.warn(f"{csv_path} not found for {im_path}")
        return

    preds = model.predict(np.expand_dims(inputs, 0))[0]
    label_map = LABELS + ["none"]
    visualize_data(inputs, [label_map[np.argmax(label)] for label in preds], f"{output_path}/{im_name}.png")
    return f"{output_path}/{im_name}.png"


article = "<div style='text-align: center;'><a href='https://nouamanetazi.me/' target='_blank'>Space by Nouamane Tazi</a><br><a href='https://keras.io/examples/vision/pointnet_segmentation' target='_blank'>Keras example by Soumik Rakshit, Sayak Paul</a></div>"

iface = gr.Interface(
    inference,  # main function
    inputs=[
        "file",
    ],
    outputs=[
        gr.outputs.Image(label="result"),  # generated image
    ],
    title="Point cloud segmentation with PointNet",
    article=article,
    examples=examples,
).launch(enable_queue=True, cache_examples=True)