app.py

import gradio as gr 
from einops import rearrange
import segmentation_models_pytorch as smp

import numpy as np
import cv2
import torch
from imutils import perspective


def midpoint(ptA, ptB):
	return ((ptA[0] + ptB[0]) * 0.5, (ptA[1] + ptB[1]) * 0.5)
# Load in image, convert to gray scale, and Otsu's threshold
kernel1 =( np.ones((5,5), dtype=np.float32))
blur_radius=0.5
kernel_sharpening = np.array([[-1,-1,-1], 
                              [-1,9,-1], 
                             [-1,-1,-1]])*(1/9)


def cca_analysis(image,predicted_mask):

    image2=np.asarray(image)
    print(image.shape)
    image = cv2.resize(predicted_mask, (image2.shape[1],image2.shape[1]), interpolation = cv2.INTER_AREA)

    image=cv2.morphologyEx(image, cv2.MORPH_OPEN, kernel1,iterations=1 )
    gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    thresh = cv2.threshold(gray, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)[1]

    labels=cv2.connectedComponents(thresh,connectivity=8)[1]       
    a=np.unique(labels)
    count2=0
    for label in a:
        if label == 0:
            continue

        # Create a mask
        mask = np.zeros(thresh.shape, dtype="uint8")
        mask[labels == label] = 255

        # Find contours and determine contour area
        cnts,hieararch = cv2.findContours(mask.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
        cnts = cnts[0]
        c_area = cv2.contourArea(cnts)
        # threshhold for tooth count
        if c_area>100:
            count2+=1
        
        rect = cv2.minAreaRect(cnts)
        box = cv2.boxPoints(rect)
        box = np.array(box, dtype="int")    
        box = perspective.order_points(box)
        color1 = (list(np.random.choice(range(150), size=3)))  
        color =[int(color1[0]), int(color1[1]), int(color1[2])]  
        cv2.drawContours(image2,[box.astype("int")],0,color,2)
        (tl,tr,br,bl)=box
        
        (tltrX,tltrY)=midpoint(tl,tr)
        (blbrX,blbrY)=midpoint(bl,br)
        # compute the midpoint between the top-left and top-right points,
        # followed by the midpoint between the top-righ and bottom-right
        (tlblX,tlblY)=midpoint(tl,bl)
        (trbrX,trbrY)=midpoint(tr,br)
        # draw the midpoints on the image
        cv2.circle(image2, (int(tltrX), int(tltrY)), 5, (255, 0, 0), -1)
        cv2.circle(image2, (int(blbrX), int(blbrY)), 5, (255, 0, 0), -1)
        cv2.circle(image2, (int(tlblX), int(tlblY)), 5, (255, 0, 0), -1)
        cv2.circle(image2, (int(trbrX), int(trbrY)), 5, (255, 0, 0), -1)
        cv2.line(image2, (int(tltrX), int(tltrY)), (int(blbrX), int(blbrY)),color, 2)
        cv2.line(image2, (int(tlblX), int(tlblY)), (int(trbrX), int(trbrY)),color, 2)
    return image2


def to_rgb(img):
    result_new=np.zeros((img.shape[1],img.shape[0],3))
    result_new[:,:,0]=img
    result_new[:,:,1]=img
    result_new[:,:,2]=img
    result_new=np.uint8(result_new*255)
    return result_new







model = smp.Unet(
    encoder_name="se_resnext50_32x4d",        
    encoder_weights=None,     
    classes=1,  
    activation="sigmoid"
)
device='cuda' if torch.cuda.is_available() else 'cpu'
map_location=torch.device(device)

state_dict = torch.load("srx50-f0.pth",map_location=map_location)["state_dict"]



model.load_state_dict(state_dict)
model.eval()

def predict(image_path,postpro): 
    output_save = "output.png"
    im = cv2.imread(image_path)
    image = cv2.resize(im, (768,768), interpolation = cv2.INTER_AREA)
    im=np.expand_dims(image,0)/255

    logits = model(torch.from_numpy(rearrange(im, 'b h w c -> b c h w')).float())
    result = logits.sigmoid()[0].detach().numpy().squeeze()

    if postpro=="Connected Components Labelling":
        result=to_rgb(result)
        result=cca_analysis(image,result)
    else:
        result=result*255

    cv2.imwrite(output_save, result)
    return image_path, output_save


title = "Deprem ML - A U-Net model of Open Cities Challange Winner 1st "

markdown=f'''

  [For the official implementation.](https://github.com/qubvel/segmentation_models.pytorch)

  
  It is running on {device}
  
'''


image_list = [
    "data/1.png",
    "data/2.png",
    "data/3.png",
    "data/4.png",
]
examples = [[image_list[0], "Connected Components Labelling"],
            [image_list[1], "Connected Components Labelling"],
            [image_list[2], "Connected Components Labelling"],
            [image_list[3], "Connected Components Labelling"]]

app = gr.Blocks()
with app:
    gr.HTML("<h1 style='text-align: center'>{}</h1>".format(title))    

    with gr.Row():
        gr.Markdown(markdown)
    with gr.Row():
        with gr.Column():
            input_video = gr.Image(type='filepath')
            cca =  gr.Dropdown(value="Connected Components Labelling", choices=["Connected Components Labelling","No Post Process"],label="Post Process")

            input_video_button = gr.Button(value="Predict")    
                    
        with gr.Column():
            output_orijinal_image = gr.Image(type='filepath')
            
        with gr.Column():
            output_mask_image =  gr.Image(type='filepath')

        
    gr.Examples(examples, inputs=[input_video,cca], outputs=[output_orijinal_image, output_mask_image], fn=predict, cache_examples=True)
    input_video_button.click(predict, inputs=[input_video,cca], outputs=[output_orijinal_image, output_mask_image])

app.launch(debug=True)