changed neural network to CNN and added some tests

source/NN/model.py

@@ -3,16 +3,26 @@ import torch
 import torch.nn.functional as F
 
 
-class Neural_Network_Model(nn.Module):
+class Conv_Neural_Network_Model(nn.Module):
-    def __init__(self, num_classes=5,hidden_layer1 = 100,hidden_layer2 = 100):
+    def __init__(self, num_classes=5,hidden_layer1 = 512,hidden_layer2 = 256):
-        super(Neural_Network_Model, self).__init__()
+        super(Conv_Neural_Network_Model, self).__init__()
-        self.fc1 = nn.Linear(150*150*3,hidden_layer1)
+        
+        self.conv1 = nn.Conv2d(in_channels=3, out_channels=32, kernel_size=3, stride=1, padding=1)
+        self.pool1 = nn.MaxPool2d(kernel_size=2, stride=2, padding=0)
+        self.conv2 = nn.Conv2d(in_channels=32, out_channels=64, kernel_size=3, stride=1, padding=1)
+        self.pool2 = nn.MaxPool2d(kernel_size=2, stride=2, padding=0)
+        self.conv3 = nn.Conv2d(in_channels=64, out_channels=128, kernel_size=3, stride=1, padding=1)
+        self.pool3 = nn.MaxPool2d(kernel_size=2, stride=2, padding=0)
+
+        self.fc1 = nn.Linear(128*12*12,hidden_layer1)
         self.fc2 = nn.Linear(hidden_layer1,hidden_layer2)
         self.out = nn.Linear(hidden_layer2,num_classes)
-# two hidden layers
 
     def forward(self, x):
-        x = x.view(-1, 150*150*3)
+        x = self.pool1(F.relu(self.conv1(x)))
+        x = self.pool2(F.relu(self.conv2(x)))
+        x = self.pool3(F.relu(self.conv3(x)))
+        x = x.view(-1, 128*12*12) #<----flattening the image
         x = self.fc1(x)
         x = torch.relu(x)
         x = self.fc2(x)

source/NN/neural_network.py

@@ -4,7 +4,7 @@ from torch.utils.data import DataLoader
 from torchvision import datasets, transforms, utils
 from torchvision.transforms import Compose, Lambda, ToTensor
 import matplotlib.pyplot as plt
-from .model import *
+from model import *
 from PIL import Image
 
 device = torch.device('cuda')
@@ -29,7 +29,7 @@ test_set = datasets.ImageFolder(root='resources/test', transform=data_transforme
 
 #function for training model
 def train(model, dataset, iter=100, batch_size=64):
-    optimizer = torch.optim.SGD(model.parameters(), lr=0.01)
+    optimizer = torch.optim.SGD(model.parameters(), lr=0.001)
     criterion = nn.NLLLoss()
     train_loader = DataLoader(dataset, batch_size=batch_size, shuffle=True)
     model.train()
@@ -56,12 +56,12 @@ def accuracy(model, dataset):
     return correct.float() / len(dataset)
 
 
-model = Neural_Network_Model()
+model = Conv_Neural_Network_Model()
 model.to(device)
 
 #loading the already saved model:
-model.load_state_dict(torch.load('model.pth'))
+# model.load_state_dict(torch.load('model.pth'))
-model.eval()
+# model.eval()
 
 #training the model:
 # train(model, train_set)
@@ -69,12 +69,10 @@ model.eval()
 # torch.save(model.state_dict(), 'model.pth')
 
 
-#TEST - loading the image and getting results:
-#testImage_path = 'resources/images/plant_photos/pexels-dxt-73640.jpg'
 
 def load_model():
-    model = Neural_Network_Model()
+    model = Conv_Neural_Network_Model()
-    model.load_state_dict(torch.load('model.pth'))
+    model.load_state_dict(torch.load('CNN_model.pth'))
     model.eval()
     return model
 
@@ -94,3 +92,23 @@ def guess_image(model, image_tensor):
 
 
 
+# image_path = 'resources/images/plant_photos/pexels-dxt-73640.jpg'
+# image_tensor = load_image(image_path)
+# prediction = guess_image(load_model(), image_tensor)
+# print(f"The predicted image is: {prediction}")
+
+#TEST - loading the image and getting results:
+testImage_path = 'resources/images/plant_photos/pexels-justus-menke-3490295-5213970.jpg'
+testImage = Image.open(testImage_path)
+testImage = data_transformer(testImage)
+testImage = testImage.unsqueeze(0)
+testImage = testImage.to(device)
+
+model.load_state_dict(torch.load('CNN_model.pth'))
+model.to(device)
+model.eval()
+
+testOutput = model(testImage)
+_, predicted = torch.max(testOutput, 1)
+predicted_class = train_set.classes[predicted.item()]
+print(f'The predicted class is: {predicted_class}')

source/main.py

@@ -3,16 +3,17 @@ import time
 import random
 import pandas as pd
 import joblib
-
 from area.constants import WIDTH, HEIGHT, TILE_SIZE
 from area.field import drawWindow
 from area.tractor import Tractor, do_actions
 from area.field import tiles, fieldX, fieldY
-from area.field import get_tile_coordinates
+from area.field import get_tile_coordinates, get_tile_index
 from ground import Dirt
 from plant import Plant
 from bfs import graphsearch, Istate, succ
 from astar import a_star
+from NN.neural_network import load_model, load_image, guess_image
+
 WIN = pygame.display.set_mode((WIDTH, HEIGHT))
 pygame.display.set_caption('Intelligent tractor')
 
@@ -23,7 +24,7 @@ def main():
     pygame.display.update()
 
 #getting coordinates of our "goal tile":
-    tile_index=127
+    tile_index = get_tile_index()
     tile_x, tile_y = get_tile_coordinates(tile_index)
     if tile_x is not None and tile_y is not None:
         print(f"Coordinates of tile {tile_index} are: ({tile_x}, {tile_y})")
@@ -128,6 +129,13 @@ def main():
         print(predykcje)
         if predykcje == 'work':
             tractor.work_on_field(tile1, d1, p1)
+
+        #guessing the image under the tile:
+        tiles[tile_index].display_photo()
+        image_path = tiles[tile_index].photo
+        image_tensor = load_image(image_path)
+        prediction = guess_image(load_model(), image_tensor)
+        print(f"The predicted image is: {prediction}")
         time.sleep(30)
         print("\n")
 

source/tile.py

@@ -2,6 +2,7 @@ import random
 import os
 import numpy as np
 import matplotlib.pyplot as plt
+from PIL import Image
 
 # path to plant images folder (used in randomize_photo function)
 folder_path = "resources/images/plant_photos"
@@ -50,9 +51,9 @@ class Tile:
             self.image = "resources/images/rock_dirt.png"
 
 
-    def displayPhoto(self):
+    def display_photo(self):
-        img = self.photo
+        image_path = self.photo
-        img = img / 2 + 0.5     # unnormalize
+        img = Image.open(image_path)
-        npimg = img.numpy()
+        plt.imshow(img)
-        plt.imshow(np.transpose(npimg, (1, 2, 0)))
+        plt.axis('off')
         plt.show()