Merge branch 'sieci-neuronowe-uczenie'

.vscode/extensions.json

@@ -0,0 +1,3 @@
+{
+  "recommendations": ["sbsnippets.pytorch-snippets"]
+}

src/import torch.py

@@ -0,0 +1,3 @@
+import torch
+x = torch.rand(5, 3)
+print(x)

src/main.py

@@ -1,4 +1,3 @@
-import sys
 import pygame
 from field import Field
 import os

src/tile.py

@@ -4,6 +4,10 @@ from kb import tractor_kb
 import pytholog as pl
 import random
 from config import TILE_SIZE, FREE_TILES
+import torch
+import torchvision.transforms as transforms
+from PIL import Image
+
 
 class Tile(pygame.sprite.Sprite):
 
@@ -26,15 +30,39 @@ class Tile(pygame.sprite.Sprite):
             self.set_type(random_vegetable)
             self.water_level = random.randint(1, 5) * 10
             self.stage = 'planted' # wczesniej to była self.faza = 'posadzono' ale stwierdzilem ze lepiej po angielsku???
+
+            classes = [
+                    "bób", "brokuł", "brukselka", "burak", "cebula",
+                    "cukinia", "dynia", "fasola", "groch", "jarmuż",
+                    "kalafior", "kalarepa", "kapusta", "marchew", 
+                    "ogórek", "papryka", "pietruszka", "pomidor",
+                    "por", "rzepa", "rzodkiewka", "sałata", "seler",
+                    "szpinak", "ziemniak"]
+
+            model = torch.load("veggies_recognition/best_model.pth")
+
+            mean = [0.5322, 0.5120, 0.3696]
+            std = [0.2487, 0.2436, 0.2531]
+
+            image_transforms = transforms.Compose([
+                transforms.Resize((224, 224)),
+                transforms.ToTensor(),
+                transforms.Normalize(torch.Tensor(mean),torch.Tensor(std))
+            ])
+
+            self.prediction = self.predict(model, image_transforms, self.image_path, classes)
+            
         else:
             if random.randint(1, 10) % 3 == 0:
                 self.set_type('water')
                 self.water_level = 100
                 self.stage = 'no_plant'
+                self.prediction = 'water'
             else:
                 self.set_type('grass')
                 self.water_level = random.randint(1, 5) * 10
                 self.stage = 'no_plant'
+                self.prediction = 'grass'
         
 
         self.rect = self.image.get_rect()
@@ -43,6 +71,15 @@ class Tile(pygame.sprite.Sprite):
 
     def draw(self, surface):
         self.tiles.draw(surface)
+    
+    def get_random_image_from_folder(self):
+        folder_path = f"veggies_recognition/veggies/testing/{self.type}"
+
+        files = [f for f in os.listdir(folder_path) if os.path.isfile(os.path.join(folder_path, f))]
+        random_file = random.choice(files)
+
+        image_path = folder_path + "/" + random_file
+        return image_path
 
     def set_type(self, type):
         self.type = type
@@ -51,9 +88,30 @@ class Tile(pygame.sprite.Sprite):
         elif self.type == 'water':
             image_path = "images/water.png"
         else:
-            image_path = f"images/vegetables/{self.type}.png"
+            #image_path = f"images/vegetables/{self.type}.png"
+            image_path = self.get_random_image_from_folder()
             if not os.path.exists(image_path):
                 image_path = "images/question.jpg"
         
+        self.image_path = image_path
         self.image = pygame.image.load(image_path).convert()
         self.image = pygame.transform.scale(self.image, (TILE_SIZE, TILE_SIZE))
+
+    def predict(self, model, image_transforms, image_path, classes):
+        model = model.eval()
+        image = Image.open(image_path)
+        image = image.convert("RGB")
+        image = image_transforms(image).float()
+        image = image.unsqueeze(0)
+
+        output = model(image)
+        _, predicted = torch.max(output.data, 1)
+
+        result = classes[predicted.item()]
+        
+
+        if result == "ziemniak":
+            result = 'marchew'
+        return result
+
+

src/tractor.py

@@ -67,7 +67,9 @@ class Tractor(pygame.sprite.Sprite):
                 neighbors.append('grass')
 
         input_data = {
-            'tile_type': self.get_current_tile().type,
+            #tutaj będzie dostawał informację ze zdjęcia
+            'tile_type': self.get_current_tile().prediction,
+            #'tile_type': self.get_current_tile().type,
             'water_level': self.get_current_tile().water_level,
             "plant_stage": self.get_current_tile().stage,
             "neighbor_N": neighbors[0],
@@ -180,6 +182,7 @@ class Tractor(pygame.sprite.Sprite):
         if (self.get_current_tile().type != 'grass' or self.get_current_tile().type == 'water'): action = 'move'
         self.prev_action = action
 
+
         match (action):
             case ('move'):
                 pass
@@ -240,9 +243,18 @@ class Tractor(pygame.sprite.Sprite):
                 self.get_current_tile().set_type('ziemniak')
         self.move_2()
         #self.action_index += 1
-        print(action)
-        return
 
+        if self.get_current_tile().type == "grass":
+            print("Co jest faktycznie: trawa")
+        elif self.get_current_tile().type == "water":
+            print("Co jest faktycznie: woda")
+        else:
+            print("Rozpoznano: ", self.get_current_tile().prediction)
+            print("Co jest faktycznie: ", self.get_current_tile().type)
+        print("\n")
+        
+        return
+    
     def log_info(self):
         # print on what tile type the tractor is on
         x = self.rect.x // TILE_SIZE