File size: 4,595 Bytes
95d3f9a
23aa32d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
0afc645
919e333
23aa32d
 
 
 
c2ae097
23aa32d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
95d3f9a
23aa32d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
919e333
23aa32d
 
 
 
 
 
 
 
 
 
 
7c9877c
23aa32d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
import gradio as gr
import random
import os
from typing import Tuple , Dict
import time 
import torch
from PIL import Image
import numpy as np
from torchvision import transforms  
import torch.nn as nn
from torch.nn.functional import relu
import requests
from io import BytesIO


class GlobalAttention(nn.Module):
    def __init__(self, num_channels):
        super(GlobalAttention, self).__init__()
        self.attention = nn.Sequential(
            nn.Conv2d(num_channels, 1, kernel_size=1),
            nn.BatchNorm2d(1),
            nn.Sigmoid()
        )

    def forward(self, x):
        attention_weights = self.attention(x)
        return x * attention_weights


class ModelWithAttention(nn.Module):
    def __init__(self, num_characters):
        super(ModelWithAttention, self).__init__()
        self.conv1 = nn.Conv2d(1, 64, kernel_size=(3, 3), padding='same')
        self.bn1 = nn.BatchNorm2d(64)
        self.pool = nn.MaxPool2d(kernel_size=(2, 2))

        self.conv2 = nn.Conv2d(64, 128, kernel_size=(3, 3), padding='same')
        self.bn2 = nn.BatchNorm2d(128)

        self.conv3 = nn.Conv2d(128, 256, kernel_size=(3, 3), padding='same')
        self.bn3 = nn.BatchNorm2d(256)

        self.conv4 = nn.Conv2d(256, 512, kernel_size=(3, 3), padding='same')
        self.bn4 = nn.BatchNorm2d(512)
        self.pool2 = nn.MaxPool2d(kernel_size=(1, 2))

        self.attention = GlobalAttention(512)  
        
        self.flatten = nn.Flatten()
        self.fc1 = nn.Linear(16384, 512)  
        self.bn5 = nn.BatchNorm1d(512)
        self.dropout1 = nn.Dropout(0.5)

        self.fc2 = nn.Linear(512, 512)
        self.bn6 = nn.BatchNorm1d(512)
        self.dropout2 = nn.Dropout(0.75)

        self.output = nn.Linear(512, num_characters)
        
        self.sm = nn.Softmax()
    def forward(self, x):
        x = self.pool(relu(self.bn1(self.conv1(x))))
        x = self.pool(relu(self.bn2(self.conv2(x))))
        x = self.pool(relu(self.bn3(self.conv3(x))))
        x = self.pool2(relu(self.bn4(self.conv4(x))))

        # x = self.attention(x)

        x = self.flatten(x)
        x = relu(self.bn5(self.fc1(x)))
        x = self.dropout1(x)

        x = relu(self.bn6(self.fc2(x)))
        x = self.dropout2(x)

        x = self.output(x)
        x = self.sm(x)
        return x



device = "cpu"
path = "Captcha(Best).pt"


from torchvision import transforms
model = ModelWithAttention(10).to(device)
model.load_state_dict(torch.load(path , map_location=torch.device('cpu')))


transform = transforms.Compose([
    transforms.ToTensor(),
    transforms.Grayscale(),
    transforms.Resize((64,64))
])


def predict(img= None , link:str = None) -> str:
  
    sizes = [
    [15,-5 , 15,27 ],
    [15,-5 , 28,40 ],
    [15,-5 , 41,53 ],
    [15,-5 , 53,65 ],
    [15,-5 , 66,78 ]]
    
    answer = ""
    
    if img != None:
        imgss = np.array((img))
        model.eval()
        for size in (sizes):
            img = imgss[size[0]:size[1], size[2]:size[3]]
            img = Image.fromarray(img)
            img = transform(img)
            img = img.unsqueeze(0)
            answer += str((torch.argmax(model(img.to(device)))).numpy())

        return answer , imgss

    if link != None:

        response = requests.get(str(link))

        if response.status_code == 200:
            imgss = np.array(Image.open(BytesIO(response.content)))
            # print("Image downloaded and converted to numpy array successfully!")

            # print(imgss.shape)
            
            model.eval()
            for size in (sizes):
                img = imgss[size[0]:size[1], size[2]:size[3]]
                img = Image.fromarray(img)
                img = transform(img)
                img = img.unsqueeze(0)
                answer += str((torch.argmax(model(img.to(device)))).cpu().numpy())
            return answer , imgss



from pathlib import Path

path = "example"

list_path = []
list_paths = os.listdir(path)
for i in list_paths:
  list_path.append(os.path.join(path , i))
# print(list_path)



import gradio as gr

title = "GIGA Captcha Solver"
description = "This Model can solve persian numbers Captcha easly"
article = "Created By A.M.Parviz <3"

# Create the Gradio demo
demo = gr.Interface(
    fn=predict,
    inputs=[gr.Image(type="pil"),
            gr.Text()],
    outputs=[
        gr.Label(num_top_classes=10, label="Predictions"),
        gr.Image()
    ],
    examples = [[img_path, ""] for img_path in list_path],
    title=title,
    description=description,
    article=article,
)
demo.launch()
            # share=True)