提交 7ad7024c 编写于 作者: M mahdi

Fixed 2D one-hot label problems in CutMix and MixUp

上级 43a61e46
......@@ -59,10 +59,17 @@ Status CutMixBatchOp::Compute(const TensorRow &input, TensorRow *output) {
// Check inputs
if (image_shape.size() != 4 || image_shape[0] != label_shape[0]) {
RETURN_STATUS_UNEXPECTED("You must make sure images are HWC or CHW and batched before calling CutMixBatch.");
RETURN_STATUS_UNEXPECTED(
"CutMixBatch: You must make sure images are HWC or CHW and batched before calling CutMixBatch.");
}
if (label_shape.size() != 2) {
RETURN_STATUS_UNEXPECTED("CutMixBatch: Label's must be in one-hot format and in a batch.");
if (!input.at(1)->type().IsInt()) {
RETURN_STATUS_UNEXPECTED("CutMixBatch: Wrong labels type. The second column (labels) must only include int types.");
}
if (label_shape.size() != 2 && label_shape.size() != 3) {
RETURN_STATUS_UNEXPECTED(
"CutMixBatch: Wrong labels shape. The second column (labels) must have a shape of NC or NLC where N is the batch "
"size, L is the number of labels in each row, "
"and C is the number of classes. labels must be in one-hot format and in a batch.");
}
if ((image_shape[1] != 1 && image_shape[1] != 3) && image_batch_format_ == ImageBatchFormat::kNCHW) {
RETURN_STATUS_UNEXPECTED("CutMixBatch: Image doesn't match the given image format.");
......@@ -84,10 +91,12 @@ Status CutMixBatchOp::Compute(const TensorRow &input, TensorRow *output) {
// Tensor holding the output labels
std::shared_ptr<Tensor> out_labels;
RETURN_IF_NOT_OK(Tensor::CreateEmpty(TensorShape(label_shape), DataType(DataType::DE_FLOAT32), &out_labels));
RETURN_IF_NOT_OK(TypeCast(std::move(input.at(1)), &out_labels, DataType(DataType::DE_FLOAT32)));
int64_t row_labels = label_shape.size() == 3 ? label_shape[1] : 1;
int64_t num_classes = label_shape.size() == 3 ? label_shape[2] : label_shape[1];
// Compute labels and images
for (int i = 0; i < image_shape[0]; i++) {
for (int64_t i = 0; i < image_shape[0]; i++) {
// Calculating lambda
// If x1 is a random variable from Gamma(a1, 1) and x2 is a random variable from Gamma(a2, 1)
// then x = x1 / (x1+x2) is a random variable from Beta(a1, a2)
......@@ -138,22 +147,29 @@ Status CutMixBatchOp::Compute(const TensorRow &input, TensorRow *output) {
}
// Compute labels
for (int j = 0; j < label_shape[1]; j++) {
if (input.at(1)->type().IsSignedInt()) {
int64_t first_value, second_value;
RETURN_IF_NOT_OK(input.at(1)->GetItemAt(&first_value, {i, j}));
RETURN_IF_NOT_OK(input.at(1)->GetItemAt(&second_value, {rand_indx[i] % label_shape[0], j}));
RETURN_IF_NOT_OK(out_labels->SetItemAt({i, j}, label_lam * first_value + (1 - label_lam) * second_value));
} else {
uint64_t first_value, second_value;
RETURN_IF_NOT_OK(input.at(1)->GetItemAt(&first_value, {i, j}));
RETURN_IF_NOT_OK(input.at(1)->GetItemAt(&second_value, {rand_indx[i] % label_shape[0], j}));
RETURN_IF_NOT_OK(out_labels->SetItemAt({i, j}, label_lam * first_value + (1 - label_lam) * second_value));
for (int64_t j = 0; j < row_labels; j++) {
for (int64_t k = 0; k < num_classes; k++) {
std::vector<int64_t> first_index = label_shape.size() == 3 ? std::vector{i, j, k} : std::vector{i, k};
std::vector<int64_t> second_index =
label_shape.size() == 3 ? std::vector{rand_indx[i], j, k} : std::vector{rand_indx[i], k};
if (input.at(1)->type().IsSignedInt()) {
int64_t first_value, second_value;
RETURN_IF_NOT_OK(input.at(1)->GetItemAt(&first_value, first_index));
RETURN_IF_NOT_OK(input.at(1)->GetItemAt(&second_value, second_index));
RETURN_IF_NOT_OK(
out_labels->SetItemAt(first_index, label_lam * first_value + (1 - label_lam) * second_value));
} else {
uint64_t first_value, second_value;
RETURN_IF_NOT_OK(input.at(1)->GetItemAt(&first_value, first_index));
RETURN_IF_NOT_OK(input.at(1)->GetItemAt(&second_value, second_index));
RETURN_IF_NOT_OK(
out_labels->SetItemAt(first_index, label_lam * first_value + (1 - label_lam) * second_value));
}
}
}
}
}
std::shared_ptr<Tensor> out_images;
RETURN_IF_NOT_OK(TensorVectorToBatchTensor(images, &out_images));
......
......@@ -38,10 +38,17 @@ Status MixUpBatchOp::Compute(const TensorRow &input, TensorRow *output) {
// Check inputs
if (image_shape.size() != 4 || image_shape[0] != label_shape[0]) {
RETURN_STATUS_UNEXPECTED("You must make sure images are HWC or CHW and batched before calling MixUpBatch.");
RETURN_STATUS_UNEXPECTED(
"MixUpBatch:You must make sure images are HWC or CHW and batched before calling MixUpBatch.");
}
if (label_shape.size() != 2) {
RETURN_STATUS_UNEXPECTED("MixUpBatch: Label's must be in one-hot format and in a batch.");
if (!input.at(1)->type().IsInt()) {
RETURN_STATUS_UNEXPECTED("MixUpBatch: Wrong labels type. The second column (labels) must only include int types.");
}
if (label_shape.size() != 2 && label_shape.size() != 3) {
RETURN_STATUS_UNEXPECTED(
"MixUpBatch: Wrong labels shape. The second column (labels) must have a shape of NC or NLC where N is the batch "
"size, L is the number of labels in each row, "
"and C is the number of classes. labels must be in one-hot format and in a batch.");
}
if ((image_shape[1] != 1 && image_shape[1] != 3) && (image_shape[3] != 1 && image_shape[3] != 3)) {
RETURN_STATUS_UNEXPECTED("MixUpBatch: Images must be in the shape of HWC or CHW.");
......@@ -65,23 +72,31 @@ Status MixUpBatchOp::Compute(const TensorRow &input, TensorRow *output) {
// Compute labels
std::shared_ptr<Tensor> out_labels;
RETURN_IF_NOT_OK(TypeCast(std::move(input.at(1)), &out_labels, DataType("float32")));
RETURN_IF_NOT_OK(TypeCast(std::move(input.at(1)), &out_labels, DataType(DataType::DE_FLOAT32)));
int64_t row_labels = label_shape.size() == 3 ? label_shape[1] : 1;
int64_t num_classes = label_shape.size() == 3 ? label_shape[2] : label_shape[1];
for (int64_t i = 0; i < label_shape[0]; i++) {
for (int64_t j = 0; j < label_shape[1]; j++) {
if (input.at(1)->type().IsSignedInt()) {
int64_t first_value, second_value;
RETURN_IF_NOT_OK(input.at(1)->GetItemAt(&first_value, {i, j}));
RETURN_IF_NOT_OK(input.at(1)->GetItemAt(&second_value, {rand_indx[i], j}));
RETURN_IF_NOT_OK(out_labels->SetItemAt({i, j}, lam * first_value + (1 - lam) * second_value));
} else {
uint64_t first_value, second_value;
RETURN_IF_NOT_OK(input.at(1)->GetItemAt(&first_value, {i, j}));
RETURN_IF_NOT_OK(input.at(1)->GetItemAt(&second_value, {rand_indx[i], j}));
RETURN_IF_NOT_OK(out_labels->SetItemAt({i, j}, lam * first_value + (1 - lam) * second_value));
for (int64_t j = 0; j < row_labels; j++) {
for (int64_t k = 0; k < num_classes; k++) {
std::vector<int64_t> first_index = label_shape.size() == 3 ? std::vector{i, j, k} : std::vector{i, k};
std::vector<int64_t> second_index =
label_shape.size() == 3 ? std::vector{rand_indx[i], j, k} : std::vector{rand_indx[i], k};
if (input.at(1)->type().IsSignedInt()) {
int64_t first_value, second_value;
RETURN_IF_NOT_OK(input.at(1)->GetItemAt(&first_value, first_index));
RETURN_IF_NOT_OK(input.at(1)->GetItemAt(&second_value, second_index));
RETURN_IF_NOT_OK(out_labels->SetItemAt(first_index, lam * first_value + (1 - lam) * second_value));
} else {
uint64_t first_value, second_value;
RETURN_IF_NOT_OK(input.at(1)->GetItemAt(&first_value, first_index));
RETURN_IF_NOT_OK(input.at(1)->GetItemAt(&second_value, second_index));
RETURN_IF_NOT_OK(out_labels->SetItemAt(first_index, lam * first_value + (1 - lam) * second_value));
}
}
}
}
// Compute images
for (int64_t i = 0; i < images.size(); i++) {
TensorShape remaining({-1});
......
......@@ -27,6 +27,7 @@ from util import save_and_check_md5, diff_mse, visualize_list, config_get_set_se
DATA_DIR = "../data/dataset/testCifar10Data"
DATA_DIR2 = "../data/dataset/testImageNetData2/train/"
DATA_DIR3 = "../data/dataset/testCelebAData/"
GENERATE_GOLDEN = False
......@@ -36,7 +37,6 @@ def test_cutmix_batch_success1(plot=False):
Test CutMixBatch op with specified alpha and prob parameters on a batch of CHW images
"""
logger.info("test_cutmix_batch_success1")
# Original Images
ds_original = ds.Cifar10Dataset(DATA_DIR, num_samples=10, shuffle=False)
ds_original = ds_original.batch(5, drop_remainder=True)
......@@ -164,6 +164,53 @@ def test_cutmix_batch_success3(plot=False):
logger.info("MSE= {}".format(str(np.mean(mse))))
def test_cutmix_batch_success4(plot=False):
"""
Test CutMixBatch on a dataset where OneHot returns a 2D vector
"""
logger.info("test_cutmix_batch_success4")
ds_original = ds.CelebADataset(DATA_DIR3, shuffle=False)
decode_op = vision.Decode()
ds_original = ds_original.map(input_columns=["image"], operations=[decode_op])
ds_original = ds_original.batch(2, drop_remainder=True)
images_original = None
for idx, (image, _) in enumerate(ds_original):
if idx == 0:
images_original = image
else:
images_original = np.append(images_original, image, axis=0)
# CutMix Images
data1 = ds.CelebADataset(dataset_dir=DATA_DIR3, shuffle=False)
decode_op = vision.Decode()
data1 = data1.map(input_columns=["image"], operations=[decode_op])
one_hot_op = data_trans.OneHot(num_classes=100)
data1 = data1.map(input_columns=["attr"], operations=one_hot_op)
cutmix_batch_op = vision.CutMixBatch(mode.ImageBatchFormat.NHWC, 0.5, 0.9)
data1 = data1.batch(2, drop_remainder=True)
data1 = data1.map(input_columns=["image", "attr"], operations=cutmix_batch_op)
images_cutmix = None
for idx, (image, _) in enumerate(data1):
if idx == 0:
images_cutmix = image
else:
images_cutmix = np.append(images_cutmix, image, axis=0)
if plot:
visualize_list(images_original, images_cutmix)
num_samples = images_original.shape[0]
mse = np.zeros(num_samples)
for i in range(num_samples):
mse[i] = diff_mse(images_cutmix[i], images_original[i])
logger.info("MSE= {}".format(str(np.mean(mse))))
def test_cutmix_batch_nhwc_md5():
"""
Test CutMixBatch on a batch of HWC images with MD5:
......@@ -368,7 +415,7 @@ def test_cutmix_batch_fail7():
images_cutmix = image
else:
images_cutmix = np.append(images_cutmix, image, axis=0)
error_message = "CutMixBatch: Label's must be in one-hot format and in a batch"
error_message = "CutMixBatch: Wrong labels shape. The second column (labels) must have a shape of NC or NLC"
assert error_message in str(error.value)
......@@ -394,6 +441,7 @@ if __name__ == "__main__":
test_cutmix_batch_success1(plot=True)
test_cutmix_batch_success2(plot=True)
test_cutmix_batch_success3(plot=True)
test_cutmix_batch_success4(plot=True)
test_cutmix_batch_nchw_md5()
test_cutmix_batch_nhwc_md5()
test_cutmix_batch_fail1()
......
......@@ -26,6 +26,7 @@ from util import save_and_check_md5, diff_mse, visualize_list, config_get_set_se
DATA_DIR = "../data/dataset/testCifar10Data"
DATA_DIR2 = "../data/dataset/testImageNetData2/train/"
DATA_DIR3 = "../data/dataset/testCelebAData/"
GENERATE_GOLDEN = False
......@@ -162,6 +163,55 @@ def test_mixup_batch_success3(plot=False):
logger.info("MSE= {}".format(str(np.mean(mse))))
def test_mixup_batch_success4(plot=False):
"""
Test MixUpBatch op on a dataset where OneHot returns a 2D vector.
Alpha parameter will be selected by default in this case
"""
logger.info("test_mixup_batch_success4")
# Original Images
ds_original = ds.CelebADataset(DATA_DIR3, shuffle=False)
decode_op = vision.Decode()
ds_original = ds_original.map(input_columns=["image"], operations=[decode_op])
ds_original = ds_original.batch(2, drop_remainder=True)
images_original = None
for idx, (image, _) in enumerate(ds_original):
if idx == 0:
images_original = image
else:
images_original = np.append(images_original, image, axis=0)
# MixUp Images
data1 = ds.CelebADataset(DATA_DIR3, shuffle=False)
decode_op = vision.Decode()
data1 = data1.map(input_columns=["image"], operations=[decode_op])
one_hot_op = data_trans.OneHot(num_classes=100)
data1 = data1.map(input_columns=["attr"], operations=one_hot_op)
mixup_batch_op = vision.MixUpBatch()
data1 = data1.batch(2, drop_remainder=True)
data1 = data1.map(input_columns=["image", "attr"], operations=mixup_batch_op)
images_mixup = np.array([])
for idx, (image, _) in enumerate(data1):
if idx == 0:
images_mixup = image
else:
images_mixup = np.append(images_mixup, image, axis=0)
if plot:
visualize_list(images_original, images_mixup)
num_samples = images_original.shape[0]
mse = np.zeros(num_samples)
for i in range(num_samples):
mse[i] = diff_mse(images_mixup[i], images_original[i])
logger.info("MSE= {}".format(str(np.mean(mse))))
def test_mixup_batch_md5():
"""
Test MixUpBatch with MD5:
......@@ -218,7 +268,7 @@ def test_mixup_batch_fail1():
images_mixup = image
else:
images_mixup = np.append(images_mixup, image, axis=0)
error_message = "You must make sure images are HWC or CHW and batch"
error_message = "You must make sure images are HWC or CHW and batched"
assert error_message in str(error.value)
......@@ -316,12 +366,50 @@ def test_mixup_batch_fail4():
assert error_message in str(error.value)
def test_mixup_batch_fail5():
"""
Test MixUpBatch Fail 5
We expect this to fail because labels are not OntHot encoded
"""
logger.info("test_mixup_batch_fail5")
# Original Images
ds_original = ds.Cifar10Dataset(DATA_DIR, num_samples=10, shuffle=False)
ds_original = ds_original.batch(5)
images_original = np.array([])
for idx, (image, _) in enumerate(ds_original):
if idx == 0:
images_original = image
else:
images_original = np.append(images_original, image, axis=0)
# MixUp Images
data1 = ds.Cifar10Dataset(DATA_DIR, num_samples=10, shuffle=False)
mixup_batch_op = vision.MixUpBatch()
data1 = data1.batch(5, drop_remainder=True)
data1 = data1.map(input_columns=["image", "label"], operations=mixup_batch_op)
with pytest.raises(RuntimeError) as error:
images_mixup = np.array([])
for idx, (image, _) in enumerate(data1):
if idx == 0:
images_mixup = image
else:
images_mixup = np.append(images_mixup, image, axis=0)
error_message = "MixUpBatch: Wrong labels shape. The second column (labels) must have a shape of NC or NLC"
assert error_message in str(error.value)
if __name__ == "__main__":
test_mixup_batch_success1(plot=True)
test_mixup_batch_success2(plot=True)
test_mixup_batch_success3(plot=True)
test_mixup_batch_success4(plot=True)
test_mixup_batch_md5()
test_mixup_batch_fail1()
test_mixup_batch_fail2()
test_mixup_batch_fail3()
test_mixup_batch_fail4()
test_mixup_batch_fail5()
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册