File size: 7,495 Bytes
7d43cc6
 
c537534
7d43cc6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d66d420
 
 
7d43cc6
d66d420
7d43cc6
 
 
 
d66d420
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
184
185
import gradio as gr
import numpy as np
import tensorflow as tf
from keras.applications.mobilenet_v2 import MobileNetV2, preprocess_input
from keras.models import Model
import matplotlib.pyplot as plt
import logging
from skimage.transform import resize
from PIL import Image
from tqdm import tqdm

class SwarmAgent:
    def __init__(self, position, velocity):
        self.position = position
        self.velocity = velocity
        self.m = np.zeros_like(position)
        self.v = np.zeros_like(position)

class SwarmNeuralNetwork:
    def __init__(self, num_agents, image_shape, target_image):
        self.image_shape = image_shape
        self.resized_shape = (64, 64, 3)
        self.agents = [SwarmAgent(self.random_position(), self.random_velocity()) for _ in range(num_agents)]
        self.target_image = self.load_target_image(target_image)
        self.generated_image = np.random.randn(*image_shape)  # Start with noise
        self.mobilenet = self.load_mobilenet_model()
        self.current_epoch = 0
        self.noise_schedule = np.linspace(0.1, 0.002, 1000)  # Noise schedule

    def random_position(self):
        return np.random.randn(*self.image_shape)  # Use Gaussian noise

    def random_velocity(self):
        return np.random.randn(*self.image_shape) * 0.01

    def load_target_image(self, img):
        img = img.resize((self.image_shape[1], self.image_shape[0]))
        img_array = np.array(img) / 127.5 - 1  # Normalize to [-1, 1]
        plt.imshow((img_array + 1) / 2)  # Convert back to [0, 1] for display
        plt.title('Target Image')
        plt.show()
        return img_array

    def resize_image(self, image):
        return resize(image, self.resized_shape, anti_aliasing=True)

    def load_mobilenet_model(self):
        mobilenet = MobileNetV2(weights='imagenet', include_top=False, input_shape=self.resized_shape)
        return Model(inputs=mobilenet.input, outputs=mobilenet.get_layer('block_13_expand_relu').output)

    def add_positional_encoding(self, image):
        h, w, c = image.shape
        pos_enc = np.zeros_like(image)
        for i in range(h):
            for j in range(w):
                pos_enc[i, j, :] = [i/h, j/w, 0]
        return image + pos_enc

    def multi_head_attention(self, agent, num_heads=4):
        attention_scores = []
        for _ in range(num_heads):
            similarity = np.exp(-np.sum((agent.position - self.target_image)**2, axis=-1))
            attention_score = similarity / np.sum(similarity)
            attention_scores.append(attention_score)
        attention = np.mean(attention_scores, axis=0)
        return np.expand_dims(attention, axis=-1)

    def multi_scale_perceptual_loss(self, agent_positions):
        target_image_resized = self.resize_image((self.target_image + 1) / 2)  # Convert to [0, 1] for MobileNet
        target_image_preprocessed = preprocess_input(target_image_resized[np.newaxis, ...] * 255)  # MobileNet expects [0, 255]
        target_features = self.mobilenet.predict(target_image_preprocessed)

        losses = []
        for agent_position in agent_positions:
            agent_image_resized = self.resize_image((agent_position + 1) / 2)
            agent_image_preprocessed = preprocess_input(agent_image_resized[np.newaxis, ...] * 255)
            agent_features = self.mobilenet.predict(agent_image_preprocessed)

            loss = np.mean((target_features - agent_features)**2)
            losses.append(1 / (1 + loss))

        return np.array(losses)

    def update_agents(self, timestep):
        noise_level = self.noise_schedule[min(timestep, len(self.noise_schedule) - 1)]
        
        for agent in self.agents:
            # Predict noise
            predicted_noise = agent.position - self.target_image
            
            # Denoise
            denoised = (agent.position - noise_level * predicted_noise) / (1 - noise_level)
            
            # Add scaled noise for next step
            agent.position = denoised + np.random.randn(*self.image_shape) * np.sqrt(noise_level)
            
            # Clip values
            agent.position = np.clip(agent.position, -1, 1)

    def generate_image(self):
        self.generated_image = np.mean([agent.position for agent in self.agents], axis=0)
        # Normalize to [0, 1] range for display
        self.generated_image = (self.generated_image + 1) / 2
        self.generated_image = np.clip(self.generated_image, 0, 1)

    def train(self, epochs):
        logging.basicConfig(filename='training.log', level=logging.INFO)

        for epoch in tqdm(range(epochs), desc="Training Epochs"):
            self.update_agents(epoch)
            self.generate_image()

            mse = np.mean(((self.generated_image * 2 - 1) - self.target_image)**2)
            logging.info(f"Epoch {epoch}, MSE: {mse}")

            if epoch % 10 == 0:
                print(f"Epoch {epoch}, MSE: {mse}")
                self.display_image(self.generated_image, title=f'Epoch {epoch}')
            self.current_epoch += 1

    def display_image(self, image, title=''):
        plt.imshow(image)
        plt.title(title)
        plt.axis('off')
        plt.show()

    def display_agent_positions(self, epoch):
        fig, ax = plt.subplots()
        positions = np.array([agent.position for agent in self.agents])
        ax.imshow(self.generated_image, extent=[0, self.image_shape[1], 0, self.image_shape[0]])
        ax.scatter(positions[:, :, 0].flatten(), positions[:, :, 1].flatten(), s=1, c='red')
        plt.title(f'Agent Positions at Epoch {epoch}')
        plt.show()

    def save_model(self, filename):
        model_state = {
            'agents': self.agents,
            'generated_image': self.generated_image,
            'current_epoch': self.current_epoch
        }
        np.save(filename, model_state)

    def load_model(self, filename):
        model_state = np.load(filename, allow_pickle=True).item()
        self.agents = model_state['agents']
        self.generated_image = model_state['generated_image']
        self.current_epoch = model_state['current_epoch']

    def generate_new_image(self, num_steps=1000):
        for agent in self.agents:
            agent.position = np.random.randn(*self.image_shape)
        
        for step in tqdm(range(num_steps), desc="Generating Image"):
            self.update_agents(num_steps - step - 1)  # Reverse order
        
        self.generate_image()
        return self.generated_image

# Gradio Interface
def train_snn(image, num_agents, epochs):
    snn = SwarmNeuralNetwork(num_agents=num_agents, image_shape=(64, 64, 3), target_image=image)
    snn.train(epochs=epochs)
    snn.save_model('snn_model.npy')
    return snn.generated_image

def generate_new_image():
    snn = SwarmNeuralNetwork(num_agents=2000, image_shape=(64, 64, 3), target_image=None)
    snn.load_model('snn_model.npy')
    new_image = snn.generate_new_image()
    return new_image

interface = gr.Interface(
    fn=train_snn,
    inputs=[
        gr.Image(type="pil", label="Upload Target Image"),
        gr.Slider(minimum=500, maximum=3000, value=2000, label="Number of Agents"),
        gr.Slider(minimum=10, maximum=200, value=100, label="Number of Epochs")
    ],
    outputs=gr.Image(type="numpy", label="Generated Image"),
    title="Swarm Neural Network Image Generation",
    description="Upload an image and set the number of agents and epochs to train the Swarm Neural Network to generate a new image."
)

interface.launch()