Разные попытки разобраться с ViT16

README.md

@@ -1,4 +1,9 @@
-### Dataset source
+## Dataset source
 
 - https://www.cs.toronto.edu/%7Ekriz/cifar.html
 - https://www.cs.toronto.edu/%7Ekriz/cifar-10-python.tar.gz (c58f30108f718f92721af3b95e74349a)
+
+## Результат
+
+Модель на 3х сверточных слоях на тестовых данных: 71% (cifar10.py)
+Модель google/vit-base-patch16-224: 10% (вероятнее всего не разаборался как корректно ее перенастроить на классификацию по 10 категориям) (ViT16.py)

ViT16.py

@@ -3,11 +3,13 @@ import torchvision.transforms as transforms
 from torchvision.datasets import CIFAR10
 from torch.utils.data import DataLoader
 from transformers import ViTImageProcessor, ViTForImageClassification, ViTFeatureExtractor
+import matplotlib.pyplot as plt
 import torch.nn.functional as F
 from sklearn.metrics import accuracy_score, classification_report
 from torch.optim import AdamW
 from torch.nn import CrossEntropyLoss
 #import os
+import numpy as np
 
 #os.environ['PYTORCH_CUDA_ALLOC_CONF'] = 'expandable_segments:True'
 
@@ -35,16 +37,18 @@ transform = transforms.Compose([
 train_dataset = CIFAR10(root='./data', train=True, download=True, transform=transform)
 test_dataset = CIFAR10(root='./data', train=False, download=True, transform=transform)
 
-# train_loader = DataLoader(train_dataset, batch_size=64, shuffle=True)
+train_loader = DataLoader(train_dataset, batch_size=64, shuffle=True)
-# test_loader = DataLoader(test_dataset, batch_size=64, shuffle=False)
+test_loader = DataLoader(test_dataset, batch_size=64, shuffle=False)
-train_loader = DataLoader(train_dataset, batch_size=32, shuffle=True)
+# train_loader = DataLoader(train_dataset, batch_size=32, shuffle=True)
-test_loader = DataLoader(test_dataset, batch_size=32, shuffle=False)
+# test_loader = DataLoader(test_dataset, batch_size=32, shuffle=False)
 
+"""
+# Дообучение модели
+# Выполняется 2 дня
 # Оптимизатор и функция потерь
 optimizer = AdamW(model.parameters(), lr=1e-4)
 criterion = CrossEntropyLoss()
 
-# Дообучение модели
 model.train()
 num_epochs = 10
 
@@ -82,6 +86,7 @@ for epoch in range(num_epochs):
 #     # Печать средней потери за эпоху
 #     avg_epoch_loss = epoch_loss / len(train_loader)
 #     print(f'Epoch {epoch+1}/{num_epochs}, Average Loss: {avg_epoch_loss}')
+"""
 
 # Функция для оценки модели
 def evaluate_model(model, data_loader, device):
@@ -111,3 +116,35 @@ accuracy, report = evaluate_model(model, test_loader, device)
 print(f'Test Accuracy: {accuracy:.4f}')
 print('Classification Report:')
 print(report)
+
+# Загрузка 25 тестовых изображений
+test_images = []
+true_labels = []
+predicted_labels = []
+
+model.eval()
+with torch.no_grad():
+    for images, labels in test_loader:
+        images = images.to(device)
+        labels = labels.to(device)
+
+        outputs = model(images).logits
+        preds = torch.argmax(outputs, dim=1)
+
+        test_images.extend(images.cpu())
+        true_labels.extend(labels.cpu())
+        predicted_labels.extend(preds.cpu())
+
+        if len(test_images) >= 25:
+            break
+
+fig, axes = plt.subplots(5, 5, figsize=(15, 15))
+axes = axes.ravel()
+
+for i in range(5 * 5):
+    axes[i].imshow(np.transpose(test_images[:25][i].cpu().numpy(), (1, 2, 0)))
+    axes[i].set_title(f'True: {train_dataset.classes[true_labels[i]]}\nPred: {train_dataset.classes[predicted_labels[i]]}')
+    axes[i].axis('off')
+
+plt.subplots_adjust(hspace=0.4)
+plt.show()

ViT16_second.py

@@ -0,0 +1,59 @@
+import torch
+from torch.utils.data import DataLoader
+from torchvision import transforms
+from transformers import ViTForImageClassification, ViTImageProcessor
+from datasets import load_dataset
+from tqdm import tqdm
+
+# Load CIFAR-10 dataset
+dataset = load_dataset('cifar10')
+
+# Define transformations for the dataset
+transform = transforms.Compose([
+    transforms.Resize((224, 224)),  # Resize images to 224x224
+    transforms.ToTensor(),
+    transforms.Normalize(mean=[0.4914, 0.4822, 0.4465], std=[0.2023, 0.1994, 0.2010])
+])
+
+# Apply transformations to the dataset
+def preprocess_function(examples):
+    examples['pixel_values'] = [transform(image.convert("RGB")) for image in examples['img']]
+    return examples
+
+encoded_dataset = dataset.with_transform(preprocess_function)
+
+# Create DataLoader for the test set
+def collate_fn(batch):
+    pixel_values = torch.stack([item['pixel_values'] for item in batch])
+    labels = torch.tensor([item['label'] for item in batch])
+    return {'pixel_values': pixel_values, 'label': labels}
+
+test_loader = DataLoader(encoded_dataset['test'], batch_size=32, shuffle=False, collate_fn=collate_fn)
+
+# Load the pre-trained ViT model
+model_name = 'google/vit-base-patch16-224'
+image_processor = ViTImageProcessor.from_pretrained(model_name)
+model = ViTForImageClassification.from_pretrained(model_name, num_labels=10, ignore_mismatched_sizes=True)
+
+# Move model to GPU if available
+device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+model.to(device)
+
+# Evaluate the model
+model.eval()
+correct = 0
+total = 0
+
+with torch.no_grad():
+    for batch in tqdm(test_loader):
+        images = batch['pixel_values'].to(device)
+        labels = batch['label'].to(device)
+
+        outputs = model(images)
+        _, predicted = torch.max(outputs.logits, 1)
+
+        total += labels.size(0)
+        correct += (predicted == labels).sum().item()
+
+accuracy = 100 * correct / total
+print(f'Accuracy of the model on the CIFAR-10 test images: {accuracy:.2f}%')

ViT16_single.py

@@ -0,0 +1,43 @@
+import torchvision.transforms as transforms
+from transformers import ViTImageProcessor, ViTForImageClassification
+from PIL import Image
+import requests
+import torch
+from torchvision.datasets import CIFAR10
+from torch.utils.data import DataLoader
+
+classes = [
+    'airplane', 'automobile', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck'
+]
+
+# url = 'http://images.cocodataset.org/val2017/000000039769.jpg'
+# image = Image.open(requests.get(url, stream=True).raw)
+
+url = 'https://www.cs.toronto.edu/~kriz/cifar-10-sample/dog10.png'
+url = 'https://www.cs.toronto.edu/~kriz/cifar-10-sample/truck10.png'
+image = Image.open(requests.get(url, stream=True).raw)
+
+test_dataset = CIFAR10(root='./data', train=False, download=True)
+test_loader = DataLoader(test_dataset, batch_size=64, shuffle=False)
+
+processor = ViTImageProcessor.from_pretrained('google/vit-base-patch16-224')
+model = ViTForImageClassification.from_pretrained('google/vit-base-patch16-224')
+
+# Изменение количества классов на 10 (для CIFAR-10)
+model.classifier = torch.nn.Linear(model.classifier.in_features, 10)
+
+# Перенос модели на GPU, если доступно
+# device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+# model.to(device)
+# model.half().to(device)  # Load in half precision
+
+inputs = processor(images=image, return_tensors="pt")
+outputs = model(**inputs)
+logits = outputs.logits
+
+class_names = ['airplane', 'automobile', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck']
+preds = torch.argmax(logits, dim=1)
+print(preds.cpu().numpy())
+# model predicts one of the 1000 ImageNet classes
+predicted_class_idx = logits.argmax(-1).item()
+print("Predicted class:", model.config.id2label[predicted_class_idx])

cifar10.py

@@ -42,16 +42,32 @@ model.compile(optimizer='adam',
               metrics=['accuracy'])
 
 # Train the model
-history = model.fit(train_images, train_labels, epochs=10,
+history = model.fit(train_images, train_labels, epochs=20,
                     validation_data=(test_images, test_labels))
 
-# Evaluate the model
+# Оценка модели
-plt.plot(history.history['accuracy'], label='accuracy')
+test_loss, test_acc = model.evaluate(train_images, train_labels, verbose=2)
-plt.plot(history.history['val_accuracy'], label = 'val_accuracy')
+print(f"Test accuracy: {test_acc:.4f}")
+
+# Построение графиков
+plt.figure(figsize=(12, 4))
+
+# График точности
+plt.subplot(1, 2, 1)
+plt.plot(history.history['accuracy'], label='train accuracy')
+plt.plot(history.history['val_accuracy'], label='val accuracy')
 plt.xlabel('Epoch')
 plt.ylabel('Accuracy')
-plt.ylim([0, 1])
 plt.legend(loc='lower right')
+plt.title('Training and Validation Accuracy')
 
-test_loss, test_acc = model.evaluate(test_images,  test_labels, verbose=2)
+# График потерь
-print(test_acc)
+plt.subplot(1, 2, 2)
+plt.plot(history.history['loss'], label='train loss')
+plt.plot(history.history['val_loss'], label='val loss')
+plt.xlabel('Epoch')
+plt.ylabel('Loss')
+plt.legend(loc='upper right')
+plt.title('Training and Validation Loss')
+
+plt.show()

cifral10_2.py

@@ -27,6 +27,7 @@ model = models.Sequential([
     layers.Conv2D(64, (3, 3), activation='relu'),
     layers.MaxPooling2D((2, 2)),
     layers.Conv2D(64, (3, 3), activation='relu'),
+
     layers.Flatten(),
     layers.Dense(64, activation='relu'),
     layers.Dense(10)

nateraw.py

@@ -0,0 +1,26 @@
+from transformers import ViTFeatureExtractor, ViTForImageClassification
+from PIL import Image
+import requests
+from tensorflow.keras import datasets, layers, models
+import matplotlib.pyplot as plt
+
+# Load the CIFAR-10 dataset
+(train_images, train_labels), (test_images, test_labels) = datasets.cifar10.load_data()
+
+# Normalize pixel values to be between 0 and 1
+train_images, test_images = train_images / 255.0, test_images / 255.0
+
+url = 'https://www.cs.toronto.edu/~kriz/cifar-10-sample/dog10.png'
+image = Image.open(requests.get(url, stream=True).raw)
+feature_extractor = ViTFeatureExtractor.from_pretrained('nateraw/vit-base-patch16-224-cifar10')
+model = ViTForImageClassification.from_pretrained('nateraw/vit-base-patch16-224-cifar10')
+inputs = feature_extractor(images=test_images[10], return_tensors="pt")
+outputs = model(**inputs)
+preds = outputs.logits.argmax(dim=1)
+
+classes = [
+    'airplane', 'automobile', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck'
+]
+print(classes[test_labels[10][0]])
+print(classes[preds[0]])
+

readme.ipynb

@@ -8,11 +8,11 @@
     "\n",
     "## Введение\n",
     "\n",
-    "Задача: Написать нейронную сеть для распознования набора данных Cifral10\n",
+    "Задача: Написать нейронную сеть для распознования набора данных Cifar10\n",
     "\n",
     "## Загрузка и предобработка данных\n",
     "\n",
-    "В качестве набора данных для обучения используется набор данных (Cifral10)[https://www.cs.toronto.edu/%7Ekriz/cifar.html]"
+    "В качестве набора данных для обучения используется набор данных (Cifar10)[https://www.cs.toronto.edu/%7Ekriz/cifar.html]"
    ]
   },
   {
@@ -171,6 +171,25 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
+    "## Результаты\n",
+    "\n",
+    "При такой модели на тестовых данных получено в районе 70% (максимум 71% на 11й эпохе). При этом тренировачные данные показывали 90% на 20й эпохе"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## google/vit-base-patch16-224\n",
+    "\n",
+    "При попытке использовать готовую модель google/vit-base-patch16-224 точность модель составила 10%. При попытке дообучить модель точность практически не изменилась. Полагаю не разборался как проверить данную модель на данных CIFAR10."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Модель: google/vit-base-patch16-224\n",
     "Test Accuracy: 0.1000\n",
     "Classification Report:\n",
     "              precision    recall  f1-score   support\n",
@@ -192,6 +211,7 @@
     "\n",
     "---\n",
     "\n",
+    "Дообученная модель google/vit-base-patch16-224 на данных Cifar10\n",
     "Test Accuracy: 0.1018\n",
     "Classification Report:\n",
     "              precision    recall  f1-score   support\n",
@@ -209,10 +229,7 @@
     "\n",
     "    accuracy                           0.10     10000\n",
     "   macro avg       0.11      0.10      0.09     10000\n",
-    "weighted avg       0.11      0.10      0.09     10000\n",
+    "weighted avg       0.11      0.10      0.09     10000"
-    "\n",
-    "\n",
-    "Process finished with exit code 0"
    ]
   }
  ],