目录

下载数据

准备输入图像的路径和目标分割掩码

一幅输入图像和相应的分割掩码是什么样子的?

准备数据集,以加载和矢量化成批数据

准备 U-Net Xception 风格模型

预留验证分割

训练模型

可视化预测


本文目标:在宠物数据集上从头开始训练的图像分割模型。

下载数据

!!wget https://www.robots.ox.ac.uk/~vgg/data/pets/data/images.tar.gz
!!wget https://www.robots.ox.ac.uk/~vgg/data/pets/data/annotations.tar.gz
!
!curl -O https://thor.robots.ox.ac.uk/datasets/pets/images.tar.gz
!curl -O https://thor.robots.ox.ac.uk/datasets/pets/annotations.tar.gz
!
!tar -xf images.tar.gz
!tar -xf annotations.tar.gz

演绎展示:

  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100  755M  100  755M    0     0  21.3M      0  0:00:35  0:00:35 --:--:-- 22.2M
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100 18.2M  100 18.2M    0     0  7977k      0  0:00:02  0:00:02 --:--:-- 7974k

准备输入图像的路径和目标分割掩码

import os

input_dir = "images/"
target_dir = "annotations/trimaps/"
img_size = (160, 160)
num_classes = 3
batch_size = 32

input_img_paths = sorted(
    [
        os.path.join(input_dir, fname)
        for fname in os.listdir(input_dir)
        if fname.endswith(".jpg")
    ]
)
target_img_paths = sorted(
    [
        os.path.join(target_dir, fname)
        for fname in os.listdir(target_dir)
        if fname.endswith(".png") and not fname.startswith(".")
    ]
)

print("Number of samples:", len(input_img_paths))

for input_path, target_path in zip(input_img_paths[:10], target_img_paths[:10]):
    print(input_path, "|", target_path)

演绎展示:

Number of samples: 7390
images/Abyssinian_1.jpg | annotations/trimaps/Abyssinian_1.png
images/Abyssinian_10.jpg | annotations/trimaps/Abyssinian_10.png
images/Abyssinian_100.jpg | annotations/trimaps/Abyssinian_100.png
images/Abyssinian_101.jpg | annotations/trimaps/Abyssinian_101.png
images/Abyssinian_102.jpg | annotations/trimaps/Abyssinian_102.png
images/Abyssinian_103.jpg | annotations/trimaps/Abyssinian_103.png
images/Abyssinian_104.jpg | annotations/trimaps/Abyssinian_104.png
images/Abyssinian_105.jpg | annotations/trimaps/Abyssinian_105.png
images/Abyssinian_106.jpg | annotations/trimaps/Abyssinian_106.png
images/Abyssinian_107.jpg | annotations/trimaps/Abyssinian_107.png

一幅输入图像和相应的分割掩码是什么样子的?

from IPython.display import Image, display
from keras.utils import load_img
from PIL import ImageOps

# Display input image #7
display(Image(filename=input_img_paths[9]))

# Display auto-contrast version of corresponding target (per-pixel categories)
img = ImageOps.autocontrast(load_img(target_img_paths[9]))
display(img)

政安晨:【Keras机器学习示例演绎】(一)—— 利用类 U-Net 架构进行图像分割-LMLPHP

政安晨:【Keras机器学习示例演绎】(一)—— 利用类 U-Net 架构进行图像分割-LMLPHP

准备数据集,以加载和矢量化成批数据

import keras
import numpy as np
from tensorflow import data as tf_data
from tensorflow import image as tf_image
from tensorflow import io as tf_io


def get_dataset(
    batch_size,
    img_size,
    input_img_paths,
    target_img_paths,
    max_dataset_len=None,
):
    """Returns a TF Dataset."""

    def load_img_masks(input_img_path, target_img_path):
        input_img = tf_io.read_file(input_img_path)
        input_img = tf_io.decode_png(input_img, channels=3)
        input_img = tf_image.resize(input_img, img_size)
        input_img = tf_image.convert_image_dtype(input_img, "float32")

        target_img = tf_io.read_file(target_img_path)
        target_img = tf_io.decode_png(target_img, channels=1)
        target_img = tf_image.resize(target_img, img_size, method="nearest")
        target_img = tf_image.convert_image_dtype(target_img, "uint8")

        # Ground truth labels are 1, 2, 3. Subtract one to make them 0, 1, 2:
        target_img -= 1
        return input_img, target_img

    # For faster debugging, limit the size of data
    if max_dataset_len:
        input_img_paths = input_img_paths[:max_dataset_len]
        target_img_paths = target_img_paths[:max_dataset_len]
    dataset = tf_data.Dataset.from_tensor_slices((input_img_paths, target_img_paths))
    dataset = dataset.map(load_img_masks, num_parallel_calls=tf_data.AUTOTUNE)
    return dataset.batch(batch_size)

准备 U-Net Xception 风格模型

from keras import layers


def get_model(img_size, num_classes):
    inputs = keras.Input(shape=img_size + (3,))

    ### [First half of the network: downsampling inputs] ###

    # Entry block
    x = layers.Conv2D(32, 3, strides=2, padding="same")(inputs)
    x = layers.BatchNormalization()(x)
    x = layers.Activation("relu")(x)

    previous_block_activation = x  # Set aside residual

    # Blocks 1, 2, 3 are identical apart from the feature depth.
    for filters in [64, 128, 256]:
        x = layers.Activation("relu")(x)
        x = layers.SeparableConv2D(filters, 3, padding="same")(x)
        x = layers.BatchNormalization()(x)

        x = layers.Activation("relu")(x)
        x = layers.SeparableConv2D(filters, 3, padding="same")(x)
        x = layers.BatchNormalization()(x)

        x = layers.MaxPooling2D(3, strides=2, padding="same")(x)

        # Project residual
        residual = layers.Conv2D(filters, 1, strides=2, padding="same")(
            previous_block_activation
        )
        x = layers.add([x, residual])  # Add back residual
        previous_block_activation = x  # Set aside next residual

    ### [Second half of the network: upsampling inputs] ###

    for filters in [256, 128, 64, 32]:
        x = layers.Activation("relu")(x)
        x = layers.Conv2DTranspose(filters, 3, padding="same")(x)
        x = layers.BatchNormalization()(x)

        x = layers.Activation("relu")(x)
        x = layers.Conv2DTranspose(filters, 3, padding="same")(x)
        x = layers.BatchNormalization()(x)

        x = layers.UpSampling2D(2)(x)

        # Project residual
        residual = layers.UpSampling2D(2)(previous_block_activation)
        residual = layers.Conv2D(filters, 1, padding="same")(residual)
        x = layers.add([x, residual])  # Add back residual
        previous_block_activation = x  # Set aside next residual

    # Add a per-pixel classification layer
    outputs = layers.Conv2D(num_classes, 3, activation="softmax", padding="same")(x)

    # Define the model
    model = keras.Model(inputs, outputs)
    return model


# Build model
model = get_model(img_size, num_classes)
model.summary()

演绎展示:

预留验证分割

import random

# Split our img paths into a training and a validation set
val_samples = 1000
random.Random(1337).shuffle(input_img_paths)
random.Random(1337).shuffle(target_img_paths)
train_input_img_paths = input_img_paths[:-val_samples]
train_target_img_paths = target_img_paths[:-val_samples]
val_input_img_paths = input_img_paths[-val_samples:]
val_target_img_paths = target_img_paths[-val_samples:]

# Instantiate dataset for each split
# Limit input files in `max_dataset_len` for faster epoch training time.
# Remove the `max_dataset_len` arg when running with full dataset.
train_dataset = get_dataset(
    batch_size,
    img_size,
    train_input_img_paths,
    train_target_img_paths,
    max_dataset_len=1000,
)
valid_dataset = get_dataset(
    batch_size, img_size, val_input_img_paths, val_target_img_paths
)

训练模型

# Configure the model for training.
# We use the "sparse" version of categorical_crossentropy
# because our target data is integers.
model.compile(
    optimizer=keras.optimizers.Adam(1e-4), loss="sparse_categorical_crossentropy"
)

callbacks = [
    keras.callbacks.ModelCheckpoint("oxford_segmentation.keras", save_best_only=True)
]

# Train the model, doing validation at the end of each epoch.
epochs = 50
model.fit(
    train_dataset,
    epochs=epochs,
    validation_data=valid_dataset,
    callbacks=callbacks,
    verbose=2,
)

演绎展示:
 

Epoch 1/50

WARNING: All log messages before absl::InitializeLog() is called are written to STDERR
I0000 00:00:1700414690.172044 2226172 device_compiler.h:187] Compiled cluster using XLA!  This line is logged at most once for the lifetime of the process.
Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 62s - 2s/step - loss: 1.6363 - val_loss: 2.2226
Epoch 2/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 94ms/step - loss: 0.9223 - val_loss: 1.8273
Epoch 3/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 82ms/step - loss: 0.7894 - val_loss: 2.0044
Epoch 4/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 83ms/step - loss: 0.7174 - val_loss: 2.3480
Epoch 5/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 82ms/step - loss: 0.6695 - val_loss: 2.7528
Epoch 6/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 83ms/step - loss: 0.6325 - val_loss: 3.1453
Epoch 7/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 84ms/step - loss: 0.6012 - val_loss: 3.5611
Epoch 8/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 87ms/step - loss: 0.5730 - val_loss: 4.0003
Epoch 9/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 85ms/step - loss: 0.5466 - val_loss: 4.4798
Epoch 10/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 86ms/step - loss: 0.5210 - val_loss: 5.0245
Epoch 11/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 87ms/step - loss: 0.4958 - val_loss: 5.5950
Epoch 12/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 87ms/step - loss: 0.4706 - val_loss: 6.1534
Epoch 13/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 85ms/step - loss: 0.4453 - val_loss: 6.6107
Epoch 14/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 83ms/step - loss: 0.4202 - val_loss: 6.8010
Epoch 15/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 84ms/step - loss: 0.3956 - val_loss: 6.6751
Epoch 16/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 83ms/step - loss: 0.3721 - val_loss: 6.0800
Epoch 17/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 84ms/step - loss: 0.3506 - val_loss: 5.1820
Epoch 18/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 82ms/step - loss: 0.3329 - val_loss: 4.0350
Epoch 19/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 4s - 114ms/step - loss: 0.3216 - val_loss: 3.0513
Epoch 20/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 94ms/step - loss: 0.3595 - val_loss: 2.2567
Epoch 21/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 100ms/step - loss: 0.4417 - val_loss: 1.5873
Epoch 22/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 101ms/step - loss: 0.3531 - val_loss: 1.5798
Epoch 23/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 96ms/step - loss: 0.3353 - val_loss: 1.5525
Epoch 24/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 95ms/step - loss: 0.3392 - val_loss: 1.4625
Epoch 25/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 95ms/step - loss: 0.3596 - val_loss: 0.8867
Epoch 26/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 94ms/step - loss: 0.3528 - val_loss: 0.8021
Epoch 27/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 92ms/step - loss: 0.3237 - val_loss: 0.7986
Epoch 28/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 89ms/step - loss: 0.3198 - val_loss: 0.8533
Epoch 29/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 84ms/step - loss: 0.3272 - val_loss: 1.0588
Epoch 30/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 88ms/step - loss: 0.3164 - val_loss: 1.1889
Epoch 31/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 85ms/step - loss: 0.2987 - val_loss: 0.9518
Epoch 32/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 87ms/step - loss: 0.2749 - val_loss: 0.9011
Epoch 33/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 84ms/step - loss: 0.2595 - val_loss: 0.8872
Epoch 34/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 87ms/step - loss: 0.2552 - val_loss: 1.0221
Epoch 35/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 82ms/step - loss: 0.2628 - val_loss: 1.1553
Epoch 36/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 85ms/step - loss: 0.2788 - val_loss: 2.1549
Epoch 37/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 94ms/step - loss: 0.2870 - val_loss: 1.6282
Epoch 38/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 89ms/step - loss: 0.2702 - val_loss: 1.3201
Epoch 39/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 91ms/step - loss: 0.2569 - val_loss: 1.2364
Epoch 40/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 106ms/step - loss: 0.2523 - val_loss: 1.3673
Epoch 41/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 86ms/step - loss: 0.2570 - val_loss: 1.3999
Epoch 42/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 87ms/step - loss: 0.2680 - val_loss: 0.9976
Epoch 43/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 83ms/step - loss: 0.2558 - val_loss: 1.0209
Epoch 44/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 85ms/step - loss: 0.2403 - val_loss: 1.3271
Epoch 45/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 83ms/step - loss: 0.2414 - val_loss: 1.1993
Epoch 46/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 84ms/step - loss: 0.2516 - val_loss: 1.0532
Epoch 47/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 83ms/step - loss: 0.2695 - val_loss: 1.1183
Epoch 48/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 87ms/step - loss: 0.2555 - val_loss: 1.0432
Epoch 49/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 82ms/step - loss: 0.2290 - val_loss: 0.9444
Epoch 50/50

Corrupt JPEG data: 240 extraneous bytes before marker 0xd9

32/32 - 3s - 83ms/step - loss: 0.1994 - val_loss: 1.2182

<keras.src.callbacks.history.History at 0x7fe01842dab0>

可视化预测

# Generate predictions for all images in the validation set

val_dataset = get_dataset(
    batch_size, img_size, val_input_img_paths, val_target_img_paths
)
val_preds = model.predict(val_dataset)


def display_mask(i):
    """Quick utility to display a model's prediction."""
    mask = np.argmax(val_preds[i], axis=-1)
    mask = np.expand_dims(mask, axis=-1)
    img = ImageOps.autocontrast(keras.utils.array_to_img(mask))
    display(img)


# Display results for validation image #10
i = 10

# Display input image
display(Image(filename=val_input_img_paths[i]))

# Display ground-truth target mask
img = ImageOps.autocontrast(load_img(val_target_img_paths[i]))
display(img)

# Display mask predicted by our model
display_mask(i)  # Note that the model only sees inputs at 150x150.

演绎展示:
 

 32/32 ━━━━━━━━━━━━━━━━━━━━ 5s 100ms/step

政安晨:【Keras机器学习示例演绎】(一)—— 利用类 U-Net 架构进行图像分割-LMLPHP

政安晨:【Keras机器学习示例演绎】(一)—— 利用类 U-Net 架构进行图像分割-LMLPHP

政安晨:【Keras机器学习示例演绎】(一)—— 利用类 U-Net 架构进行图像分割-LMLPHP


04-19 06:46