質問編集履歴

追記しました。

2020/09/16 04:18

投稿

oinari03

スコア59

test CHANGED Viewed

File without changes

test CHANGED Viewed

@@ -50,7 +50,7 @@
 実行ファイルです。
+推論をするためのファイルです。
@@ -128,7 +128,7 @@
     train_loader = torch.utils.data.DataLoader(
-        train_dataset,
+        dataset=train_dataset,
         batch_size=4,
@@ -152,7 +152,7 @@
     valid_loader = torch.utils.data.DataLoader(
-        valid_dataset,
+        dataset=valid_dataset,
         batch_size=4,
@@ -164,127 +164,267 @@
-    correct = 0
-    total = 0
     acc = []
+    num_epoch = 100
+    for epoch in range(num_epoch):
+        correct = 0 #val_acc
+        total = 0
+        with torch.no_grad():
+            for data in valid_loader:
+                inputs, labels = data[0].to(device), data[1].to(device)
+                # print("label={}".format(labels))
+                # print("inputs={}".format(inputs))
+                outputs = model(inputs).to(device)
+                # 確率
+                _, predicted = torch.max(outputs.data, 1)
+                total += labels.size(0)
+                correct += (predicted == labels).sum().item()
+            acc.append(float(correct/total))
+        print("Accuracy of the network on the 100 test images: %d/%d = %.1f" % (correct, total, 100*correct/len(valid_loader.dataset)))
+    plt.figure()
+    plt.plot(range(num_epoch), acc, color='blue',linestyle='-',label='acc')
+    plt.legend()
+    plt.xlabel('epoch')
+    plt.ylabel('acc')
+    plt.grid()
+    plt.savefig("acc.png")
+```
+### 実行結果
+一応、グラフを作ってみました。（追記）
+![イメージ説明](6f476a632313c5c68b65625788bba398.png)
+以下の部分をvalid_loader（評価用）でループを回したとき
+```
     with torch.no_grad():
         for data in valid_loader:
-            inputs, labels = data[0].to(device), data[1].to(device)
-            # print("label={}".format(labels))
-            # print("inputs={}".format(inputs))
-            outputs = model(inputs).to(device)
-            # 確率
-            _, predicted = torch.max(outputs.data, 1)
-            total += labels.size(0)
-            correct += (predicted == labels).sum().item()
-        acc.append(float(correct/total))
-    print("Accuracy of the network on the 100 test images: %d/%d = %.1f" % (correct, total, 100*correct/total))
-#グラフにしようとしたけど失敗....
-    # def plot_acc(acc):
-    #     fig, ax = plt.subplots()
-    #     epochs = np.arange(1,len(acc)+1)
-    #     ax.set_title("accuracy")
-    #     ax.plot(epochs, acc)
-    #     ax.set_xlabel("Epoch")
-    #     print("epochs: {}".format(epochs))
-    #     print("acc: {}".format(acc))
-    #     plt.savefig("acc.png")
-    # plot_acc(acc)
+```
+かなり精度はあれですが、こんな感じです。
+```
+Accuracy of the network on the 100 test images: 30/54 = 55.6
+```
+以下の部分をtrain_loader（評価用）でループを回したとき
+```
+    with torch.no_grad():
+        for data in train_loader:
+```
+100%っておかしくないか....
+```
+Accuracy of the network on the 100 test images: 129/129 = 100.0
+```
+### 追記
+前処理を行っているdataset.pyの添付ファイルをのせます。
+```
+import glob
+import os
+from PIL import Image
+from torch.utils import data as data
+from torchvision import transforms as transforms
-```
-### 実行結果
-以下の部分をvalid_loader（評価用）でループを回したとき
-```
-    with torch.no_grad():
-        for data in valid_loader:
-```
-かなり精度はあれですが、こんな感じです。
-```
-Accuracy of the network on the 100 test images: 30/54 = 55.6
-```
-以下の部分をtrain_loader（評価用）でループを回したとき
-```
-    with torch.no_grad():
-        for data in train_loader:
-```
-100%っておかしくないか....
-```
-Accuracy of the network on the 100 test images: 129/129 = 100.0
-```
+transform = transforms.Compose([
+    transforms.Resize((256, 256)),
+    transforms.ToTensor(),
+    transforms.Normalize((0.5,0.5,0.5),(0.5,0.5,0.5)),
+    # transforms.RandomHorizontalFlip(),
+    # transforms.RandomVerticalFlip(),
+    # transforms.RandomRotation(degrees=30)
+])
+class MyDatasets(data.Dataset):
+    def __init__(self, root_dir, key):
+        self.transform = transform
+        self.data = []
+        self.labels = []
+        name_to_label = {"cat": 0, "dog": 1}
+        target_dir = os.path.join(root_dir, key, "**/*")
+        for path in glob.glob(target_dir):
+            name = os.path.basename(os.path.dirname(path))
+            label = name_to_label[name]
+            self.data.append(path)
+            self.labels.append(label)
+    def __len__(self):
+        return len(self.data)
+    def __getitem__(self, index):
+        img_path = self.data[index]
+        label = self.labels[index]
+        img = Image.open(img_path).convert("RGB")
+        img = self.transform(img)
+        return img, label
+if __name__ == "__main__":
+    train_dataset = MyDatasets("./animal_dataset", "train")
+    train_dataloader = data.DataLoader(train_dataset, batch_size=4, shuffle= True)
+    for data, labels in train_dataloader:
+        print(data.shape, labels.shape)
+    datas, labels = iter(train_dataloader).next()
+    # print(datas[0].shape)
+    s=10
+    pic = transforms.ToPILImage(mode='RGB')(datas[s])
+    pic.save('./result.jpg')
+    if labels[s].numpy() == 0:
+        print("label: cat")
+    else:
+        print("label: dog")
+```
+なんかしっかり前処理できているか不安になりました。