Hugging Face's logo

Spaces:
DHEIVER
/
DR-and-Glaucoma-grading
Runtime error

App Files Community
DR-and-Glaucoma-grading
File size: 4,949 Bytes 
0c37377
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
 
import os
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import torch
import torch.nn as nn
import torch.optim as optim
from torch.utils.data import Dataset, DataLoader
from torchvision import models
from torchvision import transforms
from PIL import Image
from PIL import UnidentifiedImageError
import cv2
import segmentation_models_pytorch as sm
import gradio as gr
from aib_cdr import calculate_cdr

class Net1(nn.Module):
    def __init__(self):
        super(Net1, self).__init__()
        self.model = models.resnet50(pretrained=True)

        for param in self.model.parameters():
            param.requires_grad = False
        num_ftrs = self.model.fc.in_features
        self.model.fc = nn.Sequential(
            nn.Linear(num_ftrs, 512),
            nn.ReLU(),
            nn.Dropout(0.2),
            nn.Linear(512, 3),
            nn.Softmax(dim=1)
        )

    def forward(self, x):
        x = self.model(x)
        return x
class Net2(nn.Module):
    def __init__(self):
        super(Net2, self).__init__()
        self.model = models.resnet50(pretrained=True)

        for param in self.model.parameters():
            param.requires_grad = False
        num_ftrs = self.model.fc.in_features
        self.model.fc = nn.Sequential(
            nn.Linear(num_ftrs, 512),
            nn.ReLU(),
            nn.Dropout(0.2),
            nn.Linear(512, 2)
        )

    def forward(self, x):
        x = self.model(x)
        probas = nn.functional.softmax(x, dim=1)
        return x, probas

disc_model_path = 'models/disc_model.pth'
cup_model_path = 'models/cup_model.pth'
quality_model_path = 'models/quality_model.pth'
camtype_model_path = 'models/camtype_model.pth'
dr_model_path = 'models/dr_model.pth'

device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

disc_model = sm.Unet('resnet34', classes=2, activation=None)
disc_model.load_state_dict(torch.load(disc_model_path, map_location=device))
disc_model.to(device)
disc_model.eval()

cup_model = sm.Unet('resnet34', classes=2, activation=None)
cup_model.load_state_dict(torch.load(cup_model_path, map_location=device))
cup_model.to(device)
cup_model.eval()

quality_model = Net1().to(device)
quality_model.load_state_dict(torch.load(quality_model_path,map_location=device))
quality_model.to(device)
quality_model.eval()

camtype_model = models.resnet50(pretrained=True)
num_features = camtype_model.fc.in_features
camtype_model.fc = nn.Linear(num_features, 1)
camtype_model.load_state_dict(torch.load(camtype_model_path,map_location=device))
camtype_model.to(device)
camtype_model.eval()

dr_model = Net2().to(device)
dr_model.load_state_dict(torch.load(dr_model_path,map_location=device))
dr_model.to(device)
dr_model.eval()

transform = transforms.Compose([
    transforms.Resize((512, 512)),
    transforms.ToTensor(),  
])

def model_interface(image):
    image = Image.fromarray(image.astype('uint8'), 'RGB')
    image = image.resize((512, 512))
    image_np = np.array(image)
    image_tensor = transform(image).unsqueeze(0).to(device)

    cdr = calculate_cdr(image_tensor, disc_model, cup_model, device)
    cdr = round(cdr,2)

    quality_probs = quality_model(image_tensor)
    quality_probs = quality_probs.flatten()
    quality_dict = {"Image Quality : ACCEPTABLE": quality_probs[0].item(), "Image Quality : GOOD": quality_probs[1].item(),"Image Quality : POOR" : quality_probs[2].item()}

    camtype_pred = camtype_model(image_tensor)
    camtype_pred = camtype_pred.item()

    camtype = "Eidon" if camtype_pred == 0 else "Nidek"

    dr_grading = dr_model(image_tensor)
    dr_grading = dr_grading[1].flatten()
    dr_dict = {"DR : Negative": dr_grading[0].item(), "DR : Positive": dr_grading[1].item()}

    disc_pred = disc_model(image_tensor)
    disc_mask = (disc_pred.argmax(dim=1) > 0.5).squeeze().cpu().numpy()
    disc_contours, _ = cv2.findContours(disc_mask.astype('uint8'), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
    image_np1 = np.array(image)
    image_np1 = cv2.drawContours(image_np1, disc_contours, -1, (0, 255, 0), 2)

    cup_pred = cup_model(image_tensor)
    cup_mask = (cup_pred.argmax(dim=1) > 0.5).squeeze().cpu().numpy()
    cup_contours, _ = cv2.findContours(cup_mask.astype('uint8'), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
    image_np2 = np.array(image)
    image_np2 = cv2.drawContours(image_np2, cup_contours, -1, (0, 0, 255), 2)

    return quality_dict, camtype, dr_dict, image_np1 , image_np2,cdr

iface = gr.Interface(
    fn=model_interface,
    inputs=gr.inputs.Image(),
    outputs=[
        gr.outputs.Label(num_top_classes=3, label="Image Quality"),
        gr.outputs.Textbox(label="Camera Type"),
        gr.outputs.Label(num_top_classes=2, label="DR Grading"),
        gr.outputs.Image(type='numpy', label="Disc Mask"),
        gr.outputs.Image(type='numpy', label="Cup Mask"),
        gr.outputs.Textbox(label="Cup-to-disc Ratio")
    ]
)

iface.launch(debug=True)