More backend agnostic in cutmix and mixup file (#593)
* more backend agnostic in cutmix and mixup * date fix * reviewd comments
This commit is contained in:
Piyush Thakur
Francois Chollet
parent
7c723f242c
commit
fa22d8bf66
@ -3,7 +3,7 @@ Title: CutMix data augmentation for image classification
|
||||
Author: [Sayan Nath](https://twitter.com/sayannath2350)
|
||||
Converted to Keras Core By: [Piyush Thakur](https://github.com/cosmo3769)
|
||||
Date created: 2021/06/08
|
||||
Last modified: 2023/07/18
|
||||
Last modified: 2023/07/24
|
||||
Description: Data augmentation with CutMix for image classification on CIFAR-10.
|
||||
Accelerator: GPU
|
||||
"""
|
||||
@ -48,10 +48,16 @@ where `rx, ry` are randomly drawn from a uniform distribution with upper bound.
|
||||
|
||||
import numpy as np
|
||||
import pandas as pd
|
||||
import matplotlib.pyplot as plt
|
||||
import tensorflow as tf
|
||||
from keras_core import layers
|
||||
import keras_core as keras
|
||||
import matplotlib.pyplot as plt
|
||||
|
||||
from keras_core import layers
|
||||
|
||||
# TF imports related to tf.data preprocessing
|
||||
from tensorflow import clip_by_value
|
||||
from tensorflow import data as tf_data
|
||||
from tensorflow import image as tf_image
|
||||
from tensorflow.random import gamma as tf_random_gamma
|
||||
|
||||
keras.utils.set_random_seed(42)
|
||||
|
||||
@ -88,7 +94,7 @@ class_names = [
|
||||
## Define hyperparameters
|
||||
"""
|
||||
|
||||
AUTO = tf.data.AUTOTUNE
|
||||
AUTO = tf_data.AUTOTUNE
|
||||
BATCH_SIZE = 32
|
||||
IMG_SIZE = 32
|
||||
|
||||
@ -98,9 +104,9 @@ IMG_SIZE = 32
|
||||
|
||||
|
||||
def preprocess_image(image, label):
|
||||
image = tf.image.resize(image, (IMG_SIZE, IMG_SIZE))
|
||||
image = tf.image.convert_image_dtype(image, tf.float32) / 255.0
|
||||
label = tf.cast(label, tf.float32)
|
||||
image = tf_image.resize(image, (IMG_SIZE, IMG_SIZE))
|
||||
image = tf_image.convert_image_dtype(image, "float32") / 255.0
|
||||
label = keras.backend.cast(label, dtype="float32")
|
||||
return image, label
|
||||
|
||||
|
||||
@ -109,19 +115,19 @@ def preprocess_image(image, label):
|
||||
"""
|
||||
|
||||
train_ds_one = (
|
||||
tf.data.Dataset.from_tensor_slices((x_train, y_train))
|
||||
tf_data.Dataset.from_tensor_slices((x_train, y_train))
|
||||
.shuffle(1024)
|
||||
.map(preprocess_image, num_parallel_calls=AUTO)
|
||||
)
|
||||
train_ds_two = (
|
||||
tf.data.Dataset.from_tensor_slices((x_train, y_train))
|
||||
tf_data.Dataset.from_tensor_slices((x_train, y_train))
|
||||
.shuffle(1024)
|
||||
.map(preprocess_image, num_parallel_calls=AUTO)
|
||||
)
|
||||
|
||||
train_ds_simple = tf.data.Dataset.from_tensor_slices((x_train, y_train))
|
||||
train_ds_simple = tf_data.Dataset.from_tensor_slices((x_train, y_train))
|
||||
|
||||
test_ds = tf.data.Dataset.from_tensor_slices((x_test, y_test))
|
||||
test_ds = tf_data.Dataset.from_tensor_slices((x_test, y_test))
|
||||
|
||||
train_ds_simple = (
|
||||
train_ds_simple.map(preprocess_image, num_parallel_calls=AUTO)
|
||||
@ -130,7 +136,7 @@ train_ds_simple = (
|
||||
)
|
||||
|
||||
# Combine two shuffled datasets from the same training data.
|
||||
train_ds = tf.data.Dataset.zip((train_ds_one, train_ds_two))
|
||||
train_ds = tf_data.Dataset.zip((train_ds_one, train_ds_two))
|
||||
|
||||
test_ds = (
|
||||
test_ds.map(preprocess_image, num_parallel_calls=AUTO)
|
||||
@ -146,28 +152,29 @@ The CutMix function takes two `image` and `label` pairs to perform the augmentat
|
||||
|
||||
|
||||
def sample_beta_distribution(size, concentration_0=0.2, concentration_1=0.2):
|
||||
gamma_1_sample = tf.random.gamma(shape=[size], alpha=concentration_1)
|
||||
gamma_2_sample = tf.random.gamma(shape=[size], alpha=concentration_0)
|
||||
gamma_1_sample = tf_random_gamma(shape=[size], alpha=concentration_1)
|
||||
gamma_2_sample = tf_random_gamma(shape=[size], alpha=concentration_0)
|
||||
return gamma_1_sample / (gamma_1_sample + gamma_2_sample)
|
||||
|
||||
|
||||
@tf.function
|
||||
def get_box(lambda_value):
|
||||
cut_rat = tf.math.sqrt(1.0 - lambda_value)
|
||||
cut_rat = keras.ops.sqrt(1.0 - lambda_value)
|
||||
|
||||
cut_w = IMG_SIZE * cut_rat # rw
|
||||
cut_w = tf.cast(cut_w, tf.int32)
|
||||
cut_w = keras.backend.cast(cut_w, "int32")
|
||||
|
||||
cut_h = IMG_SIZE * cut_rat # rh
|
||||
cut_h = tf.cast(cut_h, tf.int32)
|
||||
cut_h = keras.backend.cast(cut_h, "int32")
|
||||
|
||||
cut_x = tf.random.uniform((1,), minval=0, maxval=IMG_SIZE, dtype=tf.int32) # rx
|
||||
cut_y = tf.random.uniform((1,), minval=0, maxval=IMG_SIZE, dtype=tf.int32) # ry
|
||||
cut_x = keras.random.random.uniform((1,), minval=0, maxval=IMG_SIZE) # rx
|
||||
cut_x = keras.backend.cast(cut_x, "int32")
|
||||
cut_y = keras.random.random.uniform((1,), minval=0, maxval=IMG_SIZE) # ry
|
||||
cut_y = keras.backend.cast(cut_y, "int32")
|
||||
|
||||
boundaryx1 = tf.clip_by_value(cut_x[0] - cut_w // 2, 0, IMG_SIZE)
|
||||
boundaryy1 = tf.clip_by_value(cut_y[0] - cut_h // 2, 0, IMG_SIZE)
|
||||
bbx2 = tf.clip_by_value(cut_x[0] + cut_w // 2, 0, IMG_SIZE)
|
||||
bby2 = tf.clip_by_value(cut_y[0] + cut_h // 2, 0, IMG_SIZE)
|
||||
boundaryx1 = clip_by_value(cut_x[0] - cut_w // 2, 0, IMG_SIZE)
|
||||
boundaryy1 = clip_by_value(cut_y[0] - cut_h // 2, 0, IMG_SIZE)
|
||||
bbx2 = clip_by_value(cut_x[0] + cut_w // 2, 0, IMG_SIZE)
|
||||
bby2 = clip_by_value(cut_y[0] + cut_h // 2, 0, IMG_SIZE)
|
||||
|
||||
target_h = bby2 - boundaryy1
|
||||
if target_h == 0:
|
||||
@ -180,7 +187,6 @@ def get_box(lambda_value):
|
||||
return boundaryx1, boundaryy1, target_h, target_w
|
||||
|
||||
|
||||
@tf.function
|
||||
def cutmix(train_ds_one, train_ds_two):
|
||||
(image1, label1), (image2, label2) = train_ds_one, train_ds_two
|
||||
|
||||
@ -197,19 +203,19 @@ def cutmix(train_ds_one, train_ds_two):
|
||||
boundaryx1, boundaryy1, target_h, target_w = get_box(lambda_value)
|
||||
|
||||
# Get a patch from the second image (`image2`)
|
||||
crop2 = tf.image.crop_to_bounding_box(
|
||||
crop2 = tf_image.crop_to_bounding_box(
|
||||
image2, boundaryy1, boundaryx1, target_h, target_w
|
||||
)
|
||||
# Pad the `image2` patch (`crop2`) with the same offset
|
||||
image2 = tf.image.pad_to_bounding_box(
|
||||
image2 = tf_image.pad_to_bounding_box(
|
||||
crop2, boundaryy1, boundaryx1, IMG_SIZE, IMG_SIZE
|
||||
)
|
||||
# Get a patch from the first image (`image1`)
|
||||
crop1 = tf.image.crop_to_bounding_box(
|
||||
crop1 = tf_image.crop_to_bounding_box(
|
||||
image1, boundaryy1, boundaryx1, target_h, target_w
|
||||
)
|
||||
# Pad the `image1` patch (`crop1`) with the same offset
|
||||
img1 = tf.image.pad_to_bounding_box(
|
||||
img1 = tf_image.pad_to_bounding_box(
|
||||
crop1, boundaryy1, boundaryx1, IMG_SIZE, IMG_SIZE
|
||||
)
|
||||
|
||||
@ -221,7 +227,7 @@ def cutmix(train_ds_one, train_ds_two):
|
||||
|
||||
# Adjust Lambda in accordance to the pixel ration
|
||||
lambda_value = 1 - (target_w * target_h) / (IMG_SIZE * IMG_SIZE)
|
||||
lambda_value = tf.cast(lambda_value, tf.float32)
|
||||
lambda_value = keras.backend.cast(lambda_value, "float32")
|
||||
|
||||
# Combine the labels of both images
|
||||
label = lambda_value * label1 + (1 - lambda_value) * label2
|
||||
@ -371,7 +377,7 @@ print("Test accuracy: {:.2f}%".format(test_accuracy * 100))
|
||||
|
||||
In this example, we trained our model for 15 epochs.
|
||||
In our experiment, the model with CutMix achieves a better accuracy on the CIFAR-10 dataset
|
||||
(76.92% in our experiment) compared to the model that doesn't use the augmentation (72.23%).
|
||||
(77.34% in our experiment) compared to the model that doesn't use the augmentation (66.90%).
|
||||
You may notice it takes less time to train the model with the CutMix augmentation.
|
||||
|
||||
You can experiment further with the CutMix technique by following the
|
||||
|
||||
@ -2,7 +2,7 @@
|
||||
Title: MixUp augmentation for image classification
|
||||
Author: [Sayak Paul](https://twitter.com/RisingSayak)
|
||||
Date created: 2021/03/06
|
||||
Last modified: 2021/03/06
|
||||
Last modified: 2023/07/24
|
||||
Description: Data augmentation using the mixup technique for image classification.
|
||||
Accelerator: GPU
|
||||
"""
|
||||
@ -37,10 +37,15 @@ processing, speech, and so on.
|
||||
"""
|
||||
|
||||
import numpy as np
|
||||
import tensorflow as tf
|
||||
import matplotlib.pyplot as plt
|
||||
from keras_core import layers
|
||||
import keras_core as keras
|
||||
import matplotlib.pyplot as plt
|
||||
|
||||
from keras_core import layers
|
||||
|
||||
# TF imports related to tf.data preprocessing
|
||||
from tensorflow import data as tf_data
|
||||
from tensorflow import image as tf_image
|
||||
from tensorflow.random import gamma as tf_random_gamma
|
||||
|
||||
"""
|
||||
## Prepare the dataset
|
||||
@ -53,17 +58,17 @@ be used for other classification datasets as well.
|
||||
|
||||
x_train = x_train.astype("float32") / 255.0
|
||||
x_train = np.reshape(x_train, (-1, 28, 28, 1))
|
||||
y_train = tf.one_hot(y_train, 10)
|
||||
y_train = keras.ops.one_hot(y_train, 10)
|
||||
|
||||
x_test = x_test.astype("float32") / 255.0
|
||||
x_test = np.reshape(x_test, (-1, 28, 28, 1))
|
||||
y_test = tf.one_hot(y_test, 10)
|
||||
y_test = keras.ops.one_hot(y_test, 10)
|
||||
|
||||
"""
|
||||
## Define hyperparameters
|
||||
"""
|
||||
|
||||
AUTO = tf.data.AUTOTUNE
|
||||
AUTO = tf_data.AUTOTUNE
|
||||
BATCH_SIZE = 64
|
||||
EPOCHS = 10
|
||||
|
||||
@ -77,22 +82,22 @@ x_val, y_val = x_train[:val_samples], y_train[:val_samples]
|
||||
new_x_train, new_y_train = x_train[val_samples:], y_train[val_samples:]
|
||||
|
||||
train_ds_one = (
|
||||
tf.data.Dataset.from_tensor_slices((new_x_train, new_y_train))
|
||||
tf_data.Dataset.from_tensor_slices((new_x_train, new_y_train))
|
||||
.shuffle(BATCH_SIZE * 100)
|
||||
.batch(BATCH_SIZE)
|
||||
)
|
||||
train_ds_two = (
|
||||
tf.data.Dataset.from_tensor_slices((new_x_train, new_y_train))
|
||||
tf_data.Dataset.from_tensor_slices((new_x_train, new_y_train))
|
||||
.shuffle(BATCH_SIZE * 100)
|
||||
.batch(BATCH_SIZE)
|
||||
)
|
||||
# Because we will be mixing up the images and their corresponding labels, we will be
|
||||
# combining two shuffled datasets from the same training data.
|
||||
train_ds = tf.data.Dataset.zip((train_ds_one, train_ds_two))
|
||||
train_ds = tf_data.Dataset.zip((train_ds_one, train_ds_two))
|
||||
|
||||
val_ds = tf.data.Dataset.from_tensor_slices((x_val, y_val)).batch(BATCH_SIZE)
|
||||
val_ds = tf_data.Dataset.from_tensor_slices((x_val, y_val)).batch(BATCH_SIZE)
|
||||
|
||||
test_ds = tf.data.Dataset.from_tensor_slices((x_test, y_test)).batch(BATCH_SIZE)
|
||||
test_ds = tf_data.Dataset.from_tensor_slices((x_test, y_test)).batch(BATCH_SIZE)
|
||||
|
||||
"""
|
||||
## Define the mixup technique function
|
||||
@ -105,8 +110,8 @@ the same dataset, and apply a lambda value within the [0, 1] range sampled from
|
||||
|
||||
|
||||
def sample_beta_distribution(size, concentration_0=0.2, concentration_1=0.2):
|
||||
gamma_1_sample = tf.random.gamma(shape=[size], alpha=concentration_1)
|
||||
gamma_2_sample = tf.random.gamma(shape=[size], alpha=concentration_0)
|
||||
gamma_1_sample = tf_random_gamma(shape=[size], alpha=concentration_1)
|
||||
gamma_2_sample = tf_random_gamma(shape=[size], alpha=concentration_0)
|
||||
return gamma_1_sample / (gamma_1_sample + gamma_2_sample)
|
||||
|
||||
|
||||
@ -114,12 +119,12 @@ def mix_up(ds_one, ds_two, alpha=0.2):
|
||||
# Unpack two datasets
|
||||
images_one, labels_one = ds_one
|
||||
images_two, labels_two = ds_two
|
||||
batch_size = tf.shape(images_one)[0]
|
||||
batch_size = keras.backend.shape(images_one)[0]
|
||||
|
||||
# Sample lambda and reshape it to do the mixup
|
||||
l = sample_beta_distribution(batch_size, alpha, alpha)
|
||||
x_l = tf.reshape(l, (batch_size, 1, 1, 1))
|
||||
y_l = tf.reshape(l, (batch_size, 1))
|
||||
x_l = keras.ops.reshape(l, (batch_size, 1, 1, 1))
|
||||
y_l = keras.ops.reshape(l, (batch_size, 1))
|
||||
|
||||
# Perform mixup on both images and labels by combining a pair of images/labels
|
||||
# (one from each dataset) into one image/label
|
||||
|
||||
Loading…
Reference in New Issue
Block a user