Keras 2 : examples : FixRes: 訓練/テスト解像度の不一致の修正 (翻訳/解説)

翻訳 : (株)クラスキャットセールスインフォメーション
作成日時 : 11/19/2021 (keras 2.7.0)

* 本ページは、Keras の以下のドキュメントを翻訳した上で適宜、補足説明したものです：

Code examples : Computer Vision : FixRes: Fixing train-test resolution discrepancy (Author: Sayak Paul)

* サンプルコードの動作確認はしておりますが、必要な場合には適宜、追加改変しています。
* ご自由にリンクを張って頂いてかまいませんが、sales-info@classcat.com までご一報いただけると嬉しいです。

クラスキャット人工知能研究開発支援サービス ★ 無料 Web セミナー開催中 ★

◆ クラスキャットは人工知能・テレワークに関する各種サービスを提供しております。お気軽にご相談ください :

人工知能研究開発支援
1. 人工知能研修サービス(経営者層向けオンサイト研修)
2. テクニカルコンサルティングサービス
3. 実証実験(プロトタイプ構築)
4. アプリケーションへの実装
人工知能研修サービス
PoC(概念実証)を失敗させないための支援
テレワーク & オンライン授業を支援

◆ 人工知能とビジネスをテーマに WEB セミナーを定期的に開催しています。スケジュール。

お住まいの地域に関係なく Web ブラウザからご参加頂けます。事前登録 が必要ですのでご注意ください。
ウェビナー運用には弊社製品「ClassCat® Webinar」を利用しています。

◆ お問合せ : 本件に関するお問い合わせ先は下記までお願いいたします。

株式会社クラスキャット セールス・マーケティング本部セールス・インフォメーション
E-Mail：sales-info@classcat.com ; WebSite: www.classcat.com ; Facebook

Keras 2 : examples : FixRes: 訓練/テスト解像度の不一致の修正

Description : 訓練とテストセットの間の解像度の不一致の緩和。

イントロダクション

視覚モデルを訓練してテストする際、同じ入力画像解像度を使用することは一般的な実践です。けれども、Fixing the train-test resolution discrepancy (Touvron et al.) で研究されたように、この実践は次善の性能に繋がります (最適な性能にはなりません)。データ増強は深層ニューラルネットワークの訓練プロセスの必須のパートです。視覚モデルについては、通常は訓練の間にはランダムにリサイズされたクロップを、推論の間にはセンタークロップを使用します。これは訓練と推論の間に見られるオブジェクトサイズの不一致を導入してしまいます。Touvron et al. で示されたように、この不一致を修正できれば、モデル性能を大幅にブーストできます。

このサンプルでは、この不一致を修正するために Touvron et al. により導入された FixRes テクニックを実装します。

インポート

from tensorflow import keras
from tensorflow.keras import layers
import tensorflow as tf

import tensorflow_datasets as tfds

tfds.disable_progress_bar()

import matplotlib.pyplot as plt

tf_flowers データセットのロード

train_dataset, val_dataset = tfds.load(
    "tf_flowers", split=["train[:90%]", "train[90%:]"], as_supervised=True
)

num_train = train_dataset.cardinality()
num_val = val_dataset.cardinality()
print(f"Number of training examples: {num_train}")
print(f"Number of validation examples: {num_val}")

Number of training examples: 3303
Number of validation examples: 367

データ前処理ユティリティ

3 つのデータセットを作成します :

より小さい解像度のデータセット – 128×128
より大きい解像度の 2 つのデータセット – 224×224

より大きい解像度のデータセットに様々な増強変換を適用します。

FixRes のアイデアは最初に小さい解像度のデータセットでモデルを訓練してから、それをより大きい解像度のデータセットで再調整することです。この単純で効果的なレシピは自明ではないパフォーマンス改良に繋がります。結果については元の論文を参照してください。

# Reference: https://github.com/facebookresearch/FixRes/blob/main/transforms_v2.py.

batch_size = 128
auto = tf.data.AUTOTUNE
smaller_size = 128
bigger_size = 224

size_for_resizing = int((bigger_size / smaller_size) * bigger_size)
central_crop_layer = layers.CenterCrop(bigger_size, bigger_size)


def preprocess_initial(train, image_size):
    """Initial preprocessing function for training on smaller resolution.

    For training, do random_horizontal_flip -> random_crop.
    For validation, just resize.
    No color-jittering has been used.
    """

    def _pp(image, label, train):
        if train:
            channels = image.shape[-1]
            begin, size, _ = tf.image.sample_distorted_bounding_box(
                tf.shape(image),
                tf.zeros([0, 0, 4], tf.float32),
                area_range=(0.05, 1.0),
                min_object_covered=0,
                use_image_if_no_bounding_boxes=True,
            )
            image = tf.slice(image, begin, size)

            image.set_shape([None, None, channels])
            image = tf.image.resize(image, [image_size, image_size])
            image = tf.image.random_flip_left_right(image)
        else:
            image = tf.image.resize(image, [image_size, image_size])

        return image, label

    return _pp


def preprocess_finetune(image, label, train):
    """Preprocessing function for fine-tuning on a higher resolution.

    For training, resize to a bigger resolution to maintain the ratio ->
        random_horizontal_flip -> center_crop.
    For validation, do the same without any horizontal flipping.
    No color-jittering has been used.
    """
    image = tf.image.resize(image, [size_for_resizing, size_for_resizing])
    if train:
        image = tf.image.random_flip_left_right(image)
    image = central_crop_layer(image[None, ...])[0]

    return image, label


def make_dataset(
    dataset: tf.data.Dataset,
    train: bool,
    image_size: int = smaller_size,
    fixres: bool = True,
    num_parallel_calls=auto,
):
    if image_size not in [smaller_size, bigger_size]:
        raise ValueError(f"{image_size} resolution is not supported.")

    # Determine which preprocessing function we are using.
    if image_size == smaller_size:
        preprocess_func = preprocess_initial(train, image_size)
    elif not fixres and image_size == bigger_size:
        preprocess_func = preprocess_initial(train, image_size)
    else:
        preprocess_func = preprocess_finetune

    if train:
        dataset = dataset.shuffle(batch_size * 10)

    return (
        dataset.map(
            lambda x, y: preprocess_func(x, y, train),
            num_parallel_calls=num_parallel_calls,
        )
        .batch(batch_size)
        .prefetch(num_parallel_calls)
    )

準備しているデータセットに対して増強変換がどのように変化するかに注目してください。

データセットの準備

initial_train_dataset = make_dataset(train_dataset, train=True, image_size=smaller_size)
initial_val_dataset = make_dataset(val_dataset, train=False, image_size=smaller_size)

finetune_train_dataset = make_dataset(train_dataset, train=True, image_size=bigger_size)
finetune_val_dataset = make_dataset(val_dataset, train=False, image_size=bigger_size)

vanilla_train_dataset = make_dataset(
    train_dataset, train=True, image_size=bigger_size, fixres=False
)
vanilla_val_dataset = make_dataset(
    val_dataset, train=False, image_size=bigger_size, fixres=False
)

データセットの可視化

def visualize_dataset(batch_images):
    plt.figure(figsize=(10, 10))
    for n in range(25):
        ax = plt.subplot(5, 5, n + 1)
        plt.imshow(batch_images[n].numpy().astype("int"))
        plt.axis("off")
    plt.show()

    print(f"Batch shape: {batch_images.shape}.")


# Smaller resolution.
initial_sample_images, _ = next(iter(initial_train_dataset))
visualize_dataset(initial_sample_images)

# Bigger resolution, only for fine-tuning.
finetune_sample_images, _ = next(iter(finetune_train_dataset))
visualize_dataset(finetune_sample_images)

# Bigger resolution, with the same augmentation transforms as
# the smaller resolution dataset.
vanilla_sample_images, _ = next(iter(vanilla_train_dataset))
visualize_dataset(vanilla_sample_images)

Batch shape: (128, 128, 128, 3).

Batch shape: (128, 224, 224, 3).

Batch shape: (128, 224, 224, 3).

モデル訓練ユティリティ

ResNet50V2 (He et al.) の複数のバリエーションを訓練します :

小さい解像度 (128×128) のデータセット上で。それはスクラッチから訓練されます。
次に 1 からのモデルをより大きい解像度 (224×224) のデータセットで再調整します。
別の ResNet50V2 を大きい解像度のデータセットでスクラッチから訓練します。

大きい解像度のデータセットは増強変換の観点からは異なることに注意してください。

def get_training_model(num_classes=5):
    inputs = layers.Input((None, None, 3))
    resnet_base = keras.applications.ResNet50V2(
        include_top=False, weights=None, pooling="avg"
    )
    resnet_base.trainable = True

    x = layers.Rescaling(scale=1.0 / 127.5, offset=-1)(inputs)
    x = resnet_base(x)
    outputs = layers.Dense(num_classes, activation="softmax")(x)
    return keras.Model(inputs, outputs)


def train_and_evaluate(
    model, train_ds, val_ds, epochs, learning_rate=1e-3, use_early_stopping=False
):
    optimizer = keras.optimizers.Adam(learning_rate=learning_rate)
    model.compile(
        optimizer=optimizer,
        loss="sparse_categorical_crossentropy",
        metrics=["accuracy"],
    )

    if use_early_stopping:
        es_callback = keras.callbacks.EarlyStopping(patience=5)
        callbacks = [es_callback]
    else:
        callbacks = None

    model.fit(
        train_ds, validation_data=val_ds, epochs=epochs, callbacks=callbacks,
    )

    _, accuracy = model.evaluate(val_ds)
    print(f"Top-1 accuracy on the validation set: {accuracy*100:.2f}%.")
    return model

実験 1 : 128×128 上で訓練してから 224×224 上で再調整する

epochs = 30

smaller_res_model = get_training_model()
smaller_res_model = train_and_evaluate(
    smaller_res_model, initial_train_dataset, initial_val_dataset, epochs
)

Epoch 1/30
26/26 [==============================] - 14s 226ms/step - loss: 1.6476 - accuracy: 0.4345 - val_loss: 9.8213 - val_accuracy: 0.2044
Epoch 2/30
26/26 [==============================] - 3s 123ms/step - loss: 1.1561 - accuracy: 0.5495 - val_loss: 6.5521 - val_accuracy: 0.2071
Epoch 3/30
26/26 [==============================] - 3s 123ms/step - loss: 1.0989 - accuracy: 0.5722 - val_loss: 2.6216 - val_accuracy: 0.1935
Epoch 4/30
26/26 [==============================] - 3s 122ms/step - loss: 1.0373 - accuracy: 0.5895 - val_loss: 1.9918 - val_accuracy: 0.2125
Epoch 5/30
26/26 [==============================] - 3s 122ms/step - loss: 0.9960 - accuracy: 0.6119 - val_loss: 2.8505 - val_accuracy: 0.2262
Epoch 6/30
26/26 [==============================] - 3s 122ms/step - loss: 0.9458 - accuracy: 0.6331 - val_loss: 1.8974 - val_accuracy: 0.2834
Epoch 7/30
26/26 [==============================] - 3s 122ms/step - loss: 0.8949 - accuracy: 0.6606 - val_loss: 2.1164 - val_accuracy: 0.2834
Epoch 8/30
26/26 [==============================] - 3s 122ms/step - loss: 0.8581 - accuracy: 0.6709 - val_loss: 1.8858 - val_accuracy: 0.3815
Epoch 9/30
26/26 [==============================] - 3s 123ms/step - loss: 0.8436 - accuracy: 0.6776 - val_loss: 1.5671 - val_accuracy: 0.4687
Epoch 10/30
26/26 [==============================] - 3s 123ms/step - loss: 0.8632 - accuracy: 0.6685 - val_loss: 1.5005 - val_accuracy: 0.5504
Epoch 11/30
26/26 [==============================] - 3s 123ms/step - loss: 0.8316 - accuracy: 0.6918 - val_loss: 1.1421 - val_accuracy: 0.6594
Epoch 12/30
26/26 [==============================] - 3s 123ms/step - loss: 0.7981 - accuracy: 0.6951 - val_loss: 1.2036 - val_accuracy: 0.6403
Epoch 13/30
26/26 [==============================] - 3s 122ms/step - loss: 0.8275 - accuracy: 0.6806 - val_loss: 2.2632 - val_accuracy: 0.5177
Epoch 14/30
26/26 [==============================] - 3s 122ms/step - loss: 0.8156 - accuracy: 0.6994 - val_loss: 1.1023 - val_accuracy: 0.6649
Epoch 15/30
26/26 [==============================] - 3s 122ms/step - loss: 0.7572 - accuracy: 0.7091 - val_loss: 1.6248 - val_accuracy: 0.6049
Epoch 16/30
26/26 [==============================] - 3s 123ms/step - loss: 0.7757 - accuracy: 0.7024 - val_loss: 2.0600 - val_accuracy: 0.6294
Epoch 17/30
26/26 [==============================] - 3s 122ms/step - loss: 0.7600 - accuracy: 0.7087 - val_loss: 1.5731 - val_accuracy: 0.6131
Epoch 18/30
26/26 [==============================] - 3s 122ms/step - loss: 0.7385 - accuracy: 0.7215 - val_loss: 1.8312 - val_accuracy: 0.5749
Epoch 19/30
26/26 [==============================] - 3s 122ms/step - loss: 0.7493 - accuracy: 0.7224 - val_loss: 3.0382 - val_accuracy: 0.4986
Epoch 20/30
26/26 [==============================] - 3s 122ms/step - loss: 0.7746 - accuracy: 0.7048 - val_loss: 7.8191 - val_accuracy: 0.5123
Epoch 21/30
26/26 [==============================] - 3s 123ms/step - loss: 0.7367 - accuracy: 0.7405 - val_loss: 1.9607 - val_accuracy: 0.6676
Epoch 22/30
26/26 [==============================] - 3s 122ms/step - loss: 0.6970 - accuracy: 0.7357 - val_loss: 3.1944 - val_accuracy: 0.4496
Epoch 23/30
26/26 [==============================] - 3s 122ms/step - loss: 0.7299 - accuracy: 0.7212 - val_loss: 1.4012 - val_accuracy: 0.6567
Epoch 24/30
26/26 [==============================] - 3s 122ms/step - loss: 0.6965 - accuracy: 0.7315 - val_loss: 1.9781 - val_accuracy: 0.6403
Epoch 25/30
26/26 [==============================] - 3s 124ms/step - loss: 0.6811 - accuracy: 0.7408 - val_loss: 0.9287 - val_accuracy: 0.6839
Epoch 26/30
26/26 [==============================] - 3s 123ms/step - loss: 0.6732 - accuracy: 0.7487 - val_loss: 2.9406 - val_accuracy: 0.5504
Epoch 27/30
26/26 [==============================] - 3s 122ms/step - loss: 0.6571 - accuracy: 0.7560 - val_loss: 1.6268 - val_accuracy: 0.5804
Epoch 28/30
26/26 [==============================] - 3s 122ms/step - loss: 0.6662 - accuracy: 0.7548 - val_loss: 0.9067 - val_accuracy: 0.7357
Epoch 29/30
26/26 [==============================] - 3s 122ms/step - loss: 0.6443 - accuracy: 0.7520 - val_loss: 0.7760 - val_accuracy: 0.7520
Epoch 30/30
26/26 [==============================] - 3s 122ms/step - loss: 0.6617 - accuracy: 0.7539 - val_loss: 0.6026 - val_accuracy: 0.7766
3/3 [==============================] - 0s 37ms/step - loss: 0.6026 - accuracy: 0.7766
Top-1 accuracy on the validation set: 77.66%.

(訳注 : 実験結果)

Epoch 1/30
26/26 [==============================] - 35s 353ms/step - loss: 1.4942 - accuracy: 0.4329 - val_loss: 8.4472 - val_accuracy: 0.1907
Epoch 2/30
26/26 [==============================] - 6s 213ms/step - loss: 1.1686 - accuracy: 0.5289 - val_loss: 1.7882 - val_accuracy: 0.2670
Epoch 3/30
26/26 [==============================] - 6s 212ms/step - loss: 1.1276 - accuracy: 0.5637 - val_loss: 4.7523 - val_accuracy: 0.2207
Epoch 4/30
26/26 [==============================] - 6s 211ms/step - loss: 1.1139 - accuracy: 0.5837 - val_loss: 1.5831 - val_accuracy: 0.3025
Epoch 5/30
26/26 [==============================] - 6s 213ms/step - loss: 1.0299 - accuracy: 0.6040 - val_loss: 1.9543 - val_accuracy: 0.3651
Epoch 6/30
26/26 [==============================] - 6s 212ms/step - loss: 0.9603 - accuracy: 0.6300 - val_loss: 2.1586 - val_accuracy: 0.3324
Epoch 7/30
26/26 [==============================] - 6s 213ms/step - loss: 0.9598 - accuracy: 0.6391 - val_loss: 1.3156 - val_accuracy: 0.5068
Epoch 8/30
26/26 [==============================] - 6s 216ms/step - loss: 0.9725 - accuracy: 0.6264 - val_loss: 1.9583 - val_accuracy: 0.4251
Epoch 9/30
26/26 [==============================] - 6s 213ms/step - loss: 0.9030 - accuracy: 0.6533 - val_loss: 1.6112 - val_accuracy: 0.4360
Epoch 10/30
26/26 [==============================] - 6s 213ms/step - loss: 0.8695 - accuracy: 0.6721 - val_loss: 1.2253 - val_accuracy: 0.5831
Epoch 11/30
26/26 [==============================] - 6s 211ms/step - loss: 0.8614 - accuracy: 0.6776 - val_loss: 1.0915 - val_accuracy: 0.6403
Epoch 12/30
26/26 [==============================] - 6s 212ms/step - loss: 0.8376 - accuracy: 0.6806 - val_loss: 0.8847 - val_accuracy: 0.6948
Epoch 13/30
26/26 [==============================] - 6s 211ms/step - loss: 0.8417 - accuracy: 0.6866 - val_loss: 1.6050 - val_accuracy: 0.5858
Epoch 14/30
26/26 [==============================] - 6s 213ms/step - loss: 0.8175 - accuracy: 0.6942 - val_loss: 1.4623 - val_accuracy: 0.6076
Epoch 15/30
26/26 [==============================] - 6s 213ms/step - loss: 0.8092 - accuracy: 0.6915 - val_loss: 4.5990 - val_accuracy: 0.5668
Epoch 16/30
26/26 [==============================] - 6s 213ms/step - loss: 0.7692 - accuracy: 0.7060 - val_loss: 1.5812 - val_accuracy: 0.5450
Epoch 17/30
26/26 [==============================] - 6s 212ms/step - loss: 0.7705 - accuracy: 0.7048 - val_loss: 1.1033 - val_accuracy: 0.6485
Epoch 18/30
26/26 [==============================] - 6s 211ms/step - loss: 0.7548 - accuracy: 0.7103 - val_loss: 1.2926 - val_accuracy: 0.6049
Epoch 19/30
26/26 [==============================] - 6s 211ms/step - loss: 0.7411 - accuracy: 0.7154 - val_loss: 0.9187 - val_accuracy: 0.7166
Epoch 20/30
26/26 [==============================] - 6s 213ms/step - loss: 0.7481 - accuracy: 0.7151 - val_loss: 1.5360 - val_accuracy: 0.5886
Epoch 21/30
26/26 [==============================] - 6s 211ms/step - loss: 0.7353 - accuracy: 0.7257 - val_loss: 0.8701 - val_accuracy: 0.7057
Epoch 22/30
26/26 [==============================] - 6s 212ms/step - loss: 0.7363 - accuracy: 0.7296 - val_loss: 1.3043 - val_accuracy: 0.6621
Epoch 23/30
26/26 [==============================] - 6s 212ms/step - loss: 0.7109 - accuracy: 0.7321 - val_loss: 2.6653 - val_accuracy: 0.6785
Epoch 24/30
26/26 [==============================] - 6s 211ms/step - loss: 0.7008 - accuracy: 0.7411 - val_loss: 0.6036 - val_accuracy: 0.7520
Epoch 25/30
26/26 [==============================] - 6s 211ms/step - loss: 0.6908 - accuracy: 0.7442 - val_loss: 0.9055 - val_accuracy: 0.7330
Epoch 26/30
26/26 [==============================] - 6s 211ms/step - loss: 0.6726 - accuracy: 0.7484 - val_loss: 0.9800 - val_accuracy: 0.7411
Epoch 27/30
26/26 [==============================] - 6s 215ms/step - loss: 0.6654 - accuracy: 0.7502 - val_loss: 1.0846 - val_accuracy: 0.6921
Epoch 28/30
26/26 [==============================] - 6s 212ms/step - loss: 0.6531 - accuracy: 0.7566 - val_loss: 1.0444 - val_accuracy: 0.7275
Epoch 29/30
26/26 [==============================] - 6s 216ms/step - loss: 0.6620 - accuracy: 0.7502 - val_loss: 1.2927 - val_accuracy: 0.6649
Epoch 30/30
26/26 [==============================] - 6s 214ms/step - loss: 0.6618 - accuracy: 0.7587 - val_loss: 1.0450 - val_accuracy: 0.6894
3/3 [==============================] - 0s 67ms/step - loss: 1.0450 - accuracy: 0.6894
Top-1 accuracy on the validation set: 68.94%.
CPU times: user 3min 36s, sys: 9.5 s, total: 3min 45s
Wall time: 5min 5s

最後のバッチ正規化層を除いて総ての層を凍結する

再調整のためには、2 層だけを訓練します :

最後のバッチ正規化 (Ioffe et al.) 層。
分類層。

グローバル平均プーリング層の前の最後のバッチ正規化層は活性統計情報の変化を補うために凍結解除しています。論文で示されているように、最後のバッチ正規化層の凍結解除で十分です。

Keras におけるモデルの再調整の包括的なガイドについては、このチュートリアルを参照してください。

for layer in smaller_res_model.layers[2].layers:
    layer.trainable = False

smaller_res_model.layers[2].get_layer("post_bn").trainable = True

epochs = 10

# Use a lower learning rate during fine-tuning.
bigger_res_model = train_and_evaluate(
    smaller_res_model,
    finetune_train_dataset,
    finetune_val_dataset,
    epochs,
    learning_rate=1e-4,
)

Epoch 1/10
26/26 [==============================] - 9s 201ms/step - loss: 0.7912 - accuracy: 0.7856 - val_loss: 0.6808 - val_accuracy: 0.7575
Epoch 2/10
26/26 [==============================] - 3s 115ms/step - loss: 0.7732 - accuracy: 0.7938 - val_loss: 0.7028 - val_accuracy: 0.7684
Epoch 3/10
26/26 [==============================] - 3s 115ms/step - loss: 0.7658 - accuracy: 0.7923 - val_loss: 0.7136 - val_accuracy: 0.7629
Epoch 4/10
26/26 [==============================] - 3s 115ms/step - loss: 0.7536 - accuracy: 0.7872 - val_loss: 0.7161 - val_accuracy: 0.7684
Epoch 5/10
26/26 [==============================] - 3s 115ms/step - loss: 0.7346 - accuracy: 0.7947 - val_loss: 0.7154 - val_accuracy: 0.7711
Epoch 6/10
26/26 [==============================] - 3s 115ms/step - loss: 0.7183 - accuracy: 0.7990 - val_loss: 0.7139 - val_accuracy: 0.7684
Epoch 7/10
26/26 [==============================] - 3s 116ms/step - loss: 0.7059 - accuracy: 0.7962 - val_loss: 0.7071 - val_accuracy: 0.7738
Epoch 8/10
26/26 [==============================] - 3s 115ms/step - loss: 0.6959 - accuracy: 0.7923 - val_loss: 0.7002 - val_accuracy: 0.7738
Epoch 9/10
26/26 [==============================] - 3s 116ms/step - loss: 0.6871 - accuracy: 0.8011 - val_loss: 0.6967 - val_accuracy: 0.7711
Epoch 10/10
26/26 [==============================] - 3s 116ms/step - loss: 0.6761 - accuracy: 0.8044 - val_loss: 0.6887 - val_accuracy: 0.7738
3/3 [==============================] - 0s 95ms/step - loss: 0.6887 - accuracy: 0.7738
Top-1 accuracy on the validation set: 77.38%.

Epoch 1/10
26/26 [==============================] - 15s 405ms/step - loss: 0.7293 - accuracy: 0.7663 - val_loss: 0.7613 - val_accuracy: 0.7466
Epoch 2/10
26/26 [==============================] - 9s 341ms/step - loss: 0.6958 - accuracy: 0.7654 - val_loss: 0.7010 - val_accuracy: 0.7657
Epoch 3/10
26/26 [==============================] - 9s 341ms/step - loss: 0.6887 - accuracy: 0.7729 - val_loss: 0.6589 - val_accuracy: 0.7847
Epoch 4/10
26/26 [==============================] - 9s 340ms/step - loss: 0.6240 - accuracy: 0.7781 - val_loss: 0.6232 - val_accuracy: 0.7847
Epoch 5/10
26/26 [==============================] - 9s 340ms/step - loss: 0.6314 - accuracy: 0.7829 - val_loss: 0.6033 - val_accuracy: 0.7929
Epoch 6/10
26/26 [==============================] - 9s 344ms/step - loss: 0.6280 - accuracy: 0.7944 - val_loss: 0.5856 - val_accuracy: 0.7956
Epoch 7/10
26/26 [==============================] - 9s 339ms/step - loss: 0.6094 - accuracy: 0.7911 - val_loss: 0.5780 - val_accuracy: 0.7929
Epoch 8/10
26/26 [==============================] - 9s 338ms/step - loss: 0.6000 - accuracy: 0.7884 - val_loss: 0.5683 - val_accuracy: 0.7956
Epoch 9/10
26/26 [==============================] - 9s 339ms/step - loss: 0.6183 - accuracy: 0.7953 - val_loss: 0.5645 - val_accuracy: 0.7847
Epoch 10/10
26/26 [==============================] - 9s 337ms/step - loss: 0.5987 - accuracy: 0.7978 - val_loss: 0.5567 - val_accuracy: 0.7875
3/3 [==============================] - 1s 266ms/step - loss: 0.5567 - accuracy: 0.7875
Top-1 accuracy on the validation set: 78.75%.

実験 2 : 224×224 解像度でモデルをスクラッチから訓練する

次に、大きい解像度のデータセットでスクラッチからもう一つのモデルを訓練します。このデータセットで使用される増強変換は前とは異なることを思い出してください。

epochs = 30

vanilla_bigger_res_model = get_training_model()
vanilla_bigger_res_model = train_and_evaluate(
    vanilla_bigger_res_model, vanilla_train_dataset, vanilla_val_dataset, epochs
)

Epoch 1/30
26/26 [==============================] - 15s 389ms/step - loss: 1.5339 - accuracy: 0.4569 - val_loss: 177.5233 - val_accuracy: 0.1907
Epoch 2/30
26/26 [==============================] - 8s 314ms/step - loss: 1.1472 - accuracy: 0.5483 - val_loss: 17.5804 - val_accuracy: 0.1907
Epoch 3/30
26/26 [==============================] - 8s 315ms/step - loss: 1.0708 - accuracy: 0.5792 - val_loss: 2.2719 - val_accuracy: 0.2480
Epoch 4/30
26/26 [==============================] - 8s 315ms/step - loss: 1.0225 - accuracy: 0.6170 - val_loss: 2.1274 - val_accuracy: 0.2398
Epoch 5/30
26/26 [==============================] - 8s 316ms/step - loss: 1.0001 - accuracy: 0.6206 - val_loss: 2.0375 - val_accuracy: 0.2834
Epoch 6/30
26/26 [==============================] - 8s 315ms/step - loss: 0.9602 - accuracy: 0.6355 - val_loss: 1.4412 - val_accuracy: 0.3978
Epoch 7/30
26/26 [==============================] - 8s 316ms/step - loss: 0.9418 - accuracy: 0.6461 - val_loss: 1.5257 - val_accuracy: 0.4305
Epoch 8/30
26/26 [==============================] - 8s 316ms/step - loss: 0.8911 - accuracy: 0.6649 - val_loss: 1.1530 - val_accuracy: 0.5858
Epoch 9/30
26/26 [==============================] - 8s 316ms/step - loss: 0.8834 - accuracy: 0.6694 - val_loss: 1.2026 - val_accuracy: 0.5531
Epoch 10/30
26/26 [==============================] - 8s 316ms/step - loss: 0.8752 - accuracy: 0.6724 - val_loss: 1.4917 - val_accuracy: 0.5695
Epoch 11/30
26/26 [==============================] - 8s 316ms/step - loss: 0.8690 - accuracy: 0.6594 - val_loss: 1.4115 - val_accuracy: 0.6022
Epoch 12/30
26/26 [==============================] - 8s 314ms/step - loss: 0.8586 - accuracy: 0.6761 - val_loss: 1.0692 - val_accuracy: 0.6349
Epoch 13/30
26/26 [==============================] - 8s 315ms/step - loss: 0.8120 - accuracy: 0.6894 - val_loss: 1.5233 - val_accuracy: 0.6567
Epoch 14/30
26/26 [==============================] - 8s 316ms/step - loss: 0.8275 - accuracy: 0.6857 - val_loss: 1.9079 - val_accuracy: 0.5804
Epoch 15/30
26/26 [==============================] - 8s 316ms/step - loss: 0.7624 - accuracy: 0.7127 - val_loss: 0.9543 - val_accuracy: 0.6540
Epoch 16/30
26/26 [==============================] - 8s 315ms/step - loss: 0.7595 - accuracy: 0.7266 - val_loss: 4.5757 - val_accuracy: 0.4877
Epoch 17/30
26/26 [==============================] - 8s 315ms/step - loss: 0.7577 - accuracy: 0.7154 - val_loss: 1.8411 - val_accuracy: 0.5749
Epoch 18/30
26/26 [==============================] - 8s 316ms/step - loss: 0.7596 - accuracy: 0.7163 - val_loss: 1.0660 - val_accuracy: 0.6703
Epoch 19/30
26/26 [==============================] - 8s 315ms/step - loss: 0.7492 - accuracy: 0.7160 - val_loss: 1.2462 - val_accuracy: 0.6485
Epoch 20/30
26/26 [==============================] - 8s 315ms/step - loss: 0.7269 - accuracy: 0.7330 - val_loss: 5.8287 - val_accuracy: 0.3379
Epoch 21/30
26/26 [==============================] - 8s 315ms/step - loss: 0.7193 - accuracy: 0.7275 - val_loss: 4.7058 - val_accuracy: 0.6049
Epoch 22/30
26/26 [==============================] - 8s 316ms/step - loss: 0.7251 - accuracy: 0.7318 - val_loss: 1.5608 - val_accuracy: 0.6485
Epoch 23/30
26/26 [==============================] - 8s 314ms/step - loss: 0.6888 - accuracy: 0.7466 - val_loss: 1.7914 - val_accuracy: 0.6240
Epoch 24/30
26/26 [==============================] - 8s 314ms/step - loss: 0.7051 - accuracy: 0.7339 - val_loss: 2.0918 - val_accuracy: 0.6158
Epoch 25/30
26/26 [==============================] - 8s 315ms/step - loss: 0.6920 - accuracy: 0.7454 - val_loss: 0.7284 - val_accuracy: 0.7575
Epoch 26/30
26/26 [==============================] - 8s 316ms/step - loss: 0.6502 - accuracy: 0.7523 - val_loss: 2.5474 - val_accuracy: 0.5313
Epoch 27/30
26/26 [==============================] - 8s 315ms/step - loss: 0.7101 - accuracy: 0.7330 - val_loss: 26.8117 - val_accuracy: 0.3297
Epoch 28/30
26/26 [==============================] - 8s 315ms/step - loss: 0.6632 - accuracy: 0.7548 - val_loss: 20.1011 - val_accuracy: 0.3243
Epoch 29/30
26/26 [==============================] - 8s 315ms/step - loss: 0.6682 - accuracy: 0.7505 - val_loss: 11.5872 - val_accuracy: 0.3297
Epoch 30/30
26/26 [==============================] - 8s 315ms/step - loss: 0.6758 - accuracy: 0.7514 - val_loss: 5.7229 - val_accuracy: 0.4305
3/3 [==============================] - 0s 95ms/step - loss: 5.7229 - accuracy: 0.4305
Top-1 accuracy on the validation set: 43.05%.

Epoch 1/30
26/26 [==============================] - 24s 615ms/step - loss: 1.4819 - accuracy: 0.4541 - val_loss: 204.4962 - val_accuracy: 0.2371
Epoch 2/30
26/26 [==============================] - 14s 534ms/step - loss: 1.1513 - accuracy: 0.5537 - val_loss: 8.4281 - val_accuracy: 0.2289
Epoch 3/30
26/26 [==============================] - 14s 534ms/step - loss: 1.0496 - accuracy: 0.5982 - val_loss: 3.8557 - val_accuracy: 0.1853
Epoch 4/30
26/26 [==============================] - 14s 534ms/step - loss: 1.0238 - accuracy: 0.6019 - val_loss: 2.6034 - val_accuracy: 0.2589
Epoch 5/30
26/26 [==============================] - 14s 535ms/step - loss: 0.9981 - accuracy: 0.6243 - val_loss: 2.1512 - val_accuracy: 0.3270
Epoch 6/30
26/26 [==============================] - 14s 535ms/step - loss: 0.9837 - accuracy: 0.6231 - val_loss: 2.5076 - val_accuracy: 0.2098
Epoch 7/30
26/26 [==============================] - 14s 536ms/step - loss: 0.9453 - accuracy: 0.6461 - val_loss: 2.0778 - val_accuracy: 0.3106
Epoch 8/30
26/26 [==============================] - 14s 534ms/step - loss: 0.9254 - accuracy: 0.6421 - val_loss: 1.4707 - val_accuracy: 0.5586
Epoch 9/30
26/26 [==============================] - 14s 535ms/step - loss: 0.8721 - accuracy: 0.6639 - val_loss: 1.3752 - val_accuracy: 0.5777
Epoch 10/30
26/26 [==============================] - 14s 534ms/step - loss: 0.8773 - accuracy: 0.6667 - val_loss: 1.5882 - val_accuracy: 0.5150
Epoch 11/30
26/26 [==============================] - 14s 533ms/step - loss: 0.8495 - accuracy: 0.6794 - val_loss: 1.4919 - val_accuracy: 0.5695
Epoch 12/30
26/26 [==============================] - 14s 534ms/step - loss: 0.8733 - accuracy: 0.6776 - val_loss: 1.3502 - val_accuracy: 0.4905
Epoch 13/30
26/26 [==============================] - 14s 534ms/step - loss: 0.8456 - accuracy: 0.6812 - val_loss: 0.9505 - val_accuracy: 0.6512
Epoch 14/30
26/26 [==============================] - 14s 535ms/step - loss: 0.8344 - accuracy: 0.6906 - val_loss: 3.3944 - val_accuracy: 0.4278
Epoch 15/30
26/26 [==============================] - 14s 535ms/step - loss: 0.7916 - accuracy: 0.7000 - val_loss: 1.5446 - val_accuracy: 0.6621
Epoch 16/30
26/26 [==============================] - 14s 534ms/step - loss: 0.7739 - accuracy: 0.7072 - val_loss: 1.4009 - val_accuracy: 0.6104
Epoch 17/30
26/26 [==============================] - 14s 536ms/step - loss: 0.7393 - accuracy: 0.7218 - val_loss: 1.0081 - val_accuracy: 0.6730
Epoch 18/30
26/26 [==============================] - 14s 536ms/step - loss: 0.7761 - accuracy: 0.6969 - val_loss: 1.0424 - val_accuracy: 0.6703
Epoch 19/30
26/26 [==============================] - 14s 534ms/step - loss: 0.7429 - accuracy: 0.7151 - val_loss: 0.6274 - val_accuracy: 0.7602
Epoch 20/30
26/26 [==============================] - 14s 535ms/step - loss: 0.6917 - accuracy: 0.7348 - val_loss: 1.1674 - val_accuracy: 0.6703
Epoch 21/30
26/26 [==============================] - 14s 535ms/step - loss: 0.7013 - accuracy: 0.7342 - val_loss: 0.7206 - val_accuracy: 0.7629
Epoch 22/30
26/26 [==============================] - 14s 535ms/step - loss: 0.7036 - accuracy: 0.7348 - val_loss: 0.7936 - val_accuracy: 0.7629
Epoch 23/30
26/26 [==============================] - 14s 536ms/step - loss: 0.6914 - accuracy: 0.7357 - val_loss: 1.7459 - val_accuracy: 0.6785
Epoch 24/30
26/26 [==============================] - 14s 535ms/step - loss: 0.6884 - accuracy: 0.7414 - val_loss: 1.3030 - val_accuracy: 0.7003
Epoch 25/30
26/26 [==============================] - 14s 535ms/step - loss: 0.6921 - accuracy: 0.7333 - val_loss: 1.4142 - val_accuracy: 0.6730
Epoch 26/30
26/26 [==============================] - 14s 534ms/step - loss: 0.6665 - accuracy: 0.7514 - val_loss: 0.8042 - val_accuracy: 0.7466
Epoch 27/30
26/26 [==============================] - 14s 535ms/step - loss: 0.6959 - accuracy: 0.7408 - val_loss: 2.1995 - val_accuracy: 0.6376
Epoch 28/30
26/26 [==============================] - 14s 533ms/step - loss: 0.6429 - accuracy: 0.7608 - val_loss: 1.1186 - val_accuracy: 0.6757
Epoch 29/30
26/26 [==============================] - 14s 534ms/step - loss: 0.6530 - accuracy: 0.7523 - val_loss: 1.9714 - val_accuracy: 0.6294
Epoch 30/30
26/26 [==============================] - 14s 535ms/step - loss: 0.6319 - accuracy: 0.7684 - val_loss: 1.1736 - val_accuracy: 0.7030
3/3 [==============================] - 1s 173ms/step - loss: 1.1736 - accuracy: 0.7030
Top-1 accuracy on the validation set: 70.30%.
CPU times: user 7min 48s, sys: 15.6 s, total: 8min 3s
Wall time: 10min 13s

上のセルから気付けるように、FixRes はより良いパフォーマンスをもたらします。FixRes のもう一つの利点は改善された合計訓練時間と GPU メモリ使用量の削減です。FixRes はモデル不可知ですから、パフォーマンスを潜在的にブーストするために任意の画像分類モデル上でそれを利用できます。

異なるランダムシードで同じコードを実行して集められたより多くの結果をここで見つけられます。

以上

2021年11月
月	火	水	木	金	土	日
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