app.py

import gradio as gr
from PIL import Image
# import torchvision
import os
from os.path import join
import utils
import warnings
warnings.filterwarnings("ignore")

model = utils.load_model("faster_rcnn_fold_3_150.pth")

example_path = join("example")
path = [
    join(example_path, "1.jpg"),
    join(example_path, "2.jpg")
]

def show_preds_image(image_path):
    # open image
    im = Image.open(image_path).convert("RGB")
    # transform image
    img_transformed, scale = utils.transform_img(im)
    # pred img
    # {boxes: [[int, int, int, int], ...], lables: [string, ...], scores: [int, ...] }
    pred_img = utils.predict_image(model, img_transformed, scale)

    output_img = utils.create_img_pred(image_path, pred_img)
    return output_img

inputs_image = [
    gr.components.Image(type="filepath", label="Input Image")
]

outputs_image = [
    gr.components.Image(type="numpy", label="Output Image")
]

interface_image = gr.Interface(
    fn=show_preds_image,
    inputs=inputs_image,
    outputs=outputs_image,
    title="weld Defect Detection",
    examples=path,
    cache_examples=False
).launch()


# demo = gr.Interface(fn=greet, inputs="text", outputs="text")