app.py

import os
import gradio as gr
import numpy as np
import cv2 
from PIL import Image
import torch
from inference import SegmentPredictor, DepthPredictor
from utils import generate_PCL, PCL3, point_cloud



sam = SegmentPredictor()
dpt = DepthPredictor()
red = (255,0,0)
blue = (0,0,255)
annos = []


block = gr.Blocks()
with block:
    # States
    def point_coords_empty():
        return []
    def point_labels_empty():
        return []
    raw_image = gr.Image(type='pil', visible=False)
    point_coords = gr.State(point_coords_empty)
    point_labels = gr.State(point_labels_empty)
    masks = gr.State([])
    cutout_idx = gr.State(set())

    # UI
    with gr.Column():
        with gr.Row():
            input_image = gr.Image(label='Input', type='pil', source='webcam', tool=None)
            masks_annotated_image = gr.AnnotatedImage(label='Segments', height=512)
            pcl_figure = gr.Model3D(label="3-D Reconstruction", clear_color=[1.0, 1.0, 1.0, 1.0])
            with gr.Column():
                n_samples = gr.Slider(minimum=1e3, maximum=1e6, step=1e3, default=1e3, label='Number of Samples')
                cube_size = gr.Slider(minimum=0.000001, maximum=0.001, step=0.000001, default=0.00001, label='Cube size')
        with gr.Row():
            with gr.Column(scale=1):
                with gr.Row():
                    point_label_radio = gr.Radio(label='Point Label', choices=[1,0], value=1)
                    text = gr.Textbox(label='Mask Name')
                    reset_btn = gr.Button('New Mask')
                sam_sgmt_everything_btn = gr.Button('Segment Everything!', variant = 'primary')
                sam_decode_btn = gr.Button('Predict using points!', variant = 'primary')
                depth_reconstruction_btn = gr.Button('Depth Reconstruction', variant = 'primary')
    # components
    components = {point_coords, point_labels, raw_image, masks, cutout_idx, input_image,
                  point_label_radio, text, reset_btn, sam_sgmt_everything_btn,
                  sam_decode_btn, depth_reconstruction_btn, masks_annotated_image, n_samples, cube_size}
    
    # event - init coords
    def on_reset_btn_click(raw_image):
        return raw_image, point_coords_empty(), point_labels_empty(), None, []
    reset_btn.click(on_reset_btn_click, [raw_image], [input_image, point_coords, point_labels], queue=False)

    def on_input_image_upload(input_image):
        print("encoding")
        # encode image on upload
        sam.encode(input_image)
        print("encoding done")
        return input_image, point_coords_empty(), point_labels_empty(), None
    input_image.upload(on_input_image_upload, [input_image], [raw_image, point_coords, point_labels], queue=False)

    # event - set coords
    def on_input_image_select(input_image, point_coords, point_labels, point_label_radio, evt: gr.SelectData):
        x, y = evt.index
        color = red if point_label_radio == 0 else blue
        img = np.array(input_image)
        cv2.circle(img, (x, y), 5, color, -1)
        img = Image.fromarray(img)
        point_coords.append([x,y])
        point_labels.append(point_label_radio)
        return img, point_coords, point_labels
    input_image.select(on_input_image_select, [input_image, point_coords, point_labels, point_label_radio], [input_image, point_coords, point_labels], queue=False)

    def on_click_sam_dencode_btn(inputs):
        print("inferencing")
        image = inputs[raw_image]
        generated_mask, _, _ = sam.cond_pred(pts=np.array(inputs[point_coords]), lbls=np.array(inputs[point_labels]))
        inputs[masks].append((generated_mask, inputs[text]))
        return {masks_annotated_image: (image, inputs[masks])}
    sam_decode_btn.click(on_click_sam_dencode_btn, components, [masks_annotated_image, masks, cutout_idx], queue=True)

    def on_depth_reconstruction_btn_click(inputs):
        print("depth reconstruction")
        image = inputs[input_image]
        path = dpt.generate_obj_masks(image, inputs[n_samples], inputs[cube_size])
        return {pcl_figure: path}
    depth_reconstruction_btn.click(on_depth_reconstruction_btn_click, components, [pcl_figure], queue=False)


if __name__ == '__main__':
    block.queue()
    block.launch()