Python keras.backend.abs() Examples

def rpn_loss_regr(num_anchors):
	def rpn_loss_regr_fixed_num(y_true, y_pred):
		if K.image_dim_ordering() == 'th':
			x = y_true[:, 4 * num_anchors:, :, :] - y_pred
			x_abs = K.abs(x)
			x_bool = K.less_equal(x_abs, 1.0)
			return lambda_rpn_regr * K.sum(
				y_true[:, :4 * num_anchors, :, :] * (x_bool * (0.5 * x * x) + (1 - x_bool) * (x_abs - 0.5))) / K.sum(epsilon + y_true[:, :4 * num_anchors, :, :])
		else:
			x = y_true[:, :, :, 4 * num_anchors:] - y_pred
			x_abs = K.abs(x)
			x_bool = K.cast(K.less_equal(x_abs, 1.0), tf.float32)

			return lambda_rpn_regr * K.sum(
				y_true[:, :, :, :4 * num_anchors] * (x_bool * (0.5 * x * x) + (1 - x_bool) * (x_abs - 0.5))) / K.sum(epsilon + y_true[:, :, :, :4 * num_anchors])

	return rpn_loss_regr_fixed_num 

def rpn_loss_regr(num_anchors):
	def rpn_loss_regr_fixed_num(y_true, y_pred):
		if K.image_dim_ordering() == 'th':
			x = y_true[:, 4 * num_anchors:, :, :] - y_pred
			x_abs = K.abs(x)
			x_bool = K.less_equal(x_abs, 1.0)
			return lambda_rpn_regr * K.sum(
				y_true[:, :4 * num_anchors, :, :] * (x_bool * (0.5 * x * x) + (1 - x_bool) * (x_abs - 0.5))) / K.sum(epsilon + y_true[:, :4 * num_anchors, :, :])
		else:
			x = y_true[:, :, :, 4 * num_anchors:] - y_pred
			x_abs = K.abs(x)
			x_bool = K.cast(K.less_equal(x_abs, 1.0), tf.float32)

			return lambda_rpn_regr * K.sum(
				y_true[:, :, :, :4 * num_anchors] * (x_bool * (0.5 * x * x) + (1 - x_bool) * (x_abs - 0.5))) / K.sum(epsilon + y_true[:, :, :, :4 * num_anchors])

	return rpn_loss_regr_fixed_num 

def count(audio, model, scaler):
    # compute STFT
    X = np.abs(librosa.stft(audio, n_fft=400, hop_length=160)).T

    # apply global (featurewise) standardization to mean1, var0
    X = scaler.transform(X)

    # cut to input shape length (500 frames x 201 STFT bins)
    X = X[:500, :]

    # apply l2 normalization
    Theta = np.linalg.norm(X, axis=1) + eps
    X /= np.mean(Theta)

    # add sample dimension
    X = X[np.newaxis, ...]

    if len(model.input_shape) == 4:
        X = X[:, np.newaxis, ...]

    ys = model.predict(X, verbose=0)
    return np.argmax(ys, axis=1)[0] 

def call(self, x, mask=None):
        # ensure the the right part is always to the right of the left
        t_right_actual = self.t_left + K.abs(self.t_right)

        if K.backend() == 'theano':
            t_left = K.pattern_broadcast(self.t_left, self.param_broadcast)
            a_left = K.pattern_broadcast(self.a_left, self.param_broadcast)
            a_right = K.pattern_broadcast(self.a_right, self.param_broadcast)
            t_right_actual = K.pattern_broadcast(t_right_actual,
                                                 self.param_broadcast)
        else:
            t_left = self.t_left
            a_left = self.a_left
            a_right = self.a_right

        y_left_and_center = t_left + K.relu(x - t_left,
                                            a_left,
                                            t_right_actual - t_left)
        y_right = K.relu(x - t_right_actual) * a_right
        return y_left_and_center + y_right 

def rpn_loss_regr(num_anchors):
	def rpn_loss_regr_fixed_num(y_true, y_pred):
		if K.image_dim_ordering() == 'th':
			x = y_true[:, 4 * num_anchors:, :, :] - y_pred
			x_abs = K.abs(x)
			x_bool = K.less_equal(x_abs, 1.0)
			return lambda_rpn_regr * K.sum(
				y_true[:, :4 * num_anchors, :, :] * (x_bool * (0.5 * x * x) + (1 - x_bool) * (x_abs - 0.5))) / K.sum(epsilon + y_true[:, :4 * num_anchors, :, :])
		else:
			x = y_true[:, :, :, 4 * num_anchors:] - y_pred
			x_abs = K.abs(x)
			x_bool = K.cast(K.less_equal(x_abs, 1.0), tf.float32)

			return lambda_rpn_regr * K.sum(
				y_true[:, :, :, :4 * num_anchors] * (x_bool * (0.5 * x * x) + (1 - x_bool) * (x_abs - 0.5))) / K.sum(epsilon + y_true[:, :, :, :4 * num_anchors])

	return rpn_loss_regr_fixed_num 

def rpn_loss_regr(num_anchors):
	def rpn_loss_regr_fixed_num(y_true, y_pred):
		if K.image_dim_ordering() == 'th':
			x = y_true[:, 4 * num_anchors:, :, :] - y_pred
			x_abs = K.abs(x)
			x_bool = K.less_equal(x_abs, 1.0)
			return lambda_rpn_regr * K.sum(
				y_true[:, :4 * num_anchors, :, :] * (x_bool * (0.5 * x * x) + (1 - x_bool) * (x_abs - 0.5))) / K.sum(epsilon + y_true[:, :4 * num_anchors, :, :])
		else:
			x = y_true[:, :, :, 4 * num_anchors:] - y_pred
			x_abs = K.abs(x)
			x_bool = K.cast(K.less_equal(x_abs, 1.0), tf.float32)

			return lambda_rpn_regr * K.sum(
				y_true[:, :, :, :4 * num_anchors] * (x_bool * (0.5 * x * x) + (1 - x_bool) * (x_abs - 0.5))) / K.sum(epsilon + y_true[:, :, :, :4 * num_anchors])

	return rpn_loss_regr_fixed_num 

def optimizer(self):
        a = K.placeholder(shape=(None, ), dtype='int32')
        y = K.placeholder(shape=(None, ), dtype='float32')

        py_x = self.model.output

        a_one_hot = K.one_hot(a, self.action_size)
        q_value = K.sum(py_x * a_one_hot, axis=1)
        error = K.abs(y - q_value)

        quadratic_part = K.clip(error, 0.0, 1.0)
        linear_part = error - quadratic_part
        loss = K.mean(0.5 * K.square(quadratic_part) + linear_part)

        optimizer = RMSprop(lr=0.00025, epsilon=0.01)
        updates = optimizer.get_updates(self.model.trainable_weights, [], loss)
        train = K.function([self.model.input, a, y], [loss], updates=updates)

        return train

    # approximate Q function using Convolution Neural Network
    # state is input and Q Value of each action is output of network
    # dueling network's Q Value is sum of advantages and state value 

def optimizer(self):
        a = K.placeholder(shape=(None,), dtype='int32')
        y = K.placeholder(shape=(None,), dtype='float32')

        py_x = self.model.output

        a_one_hot = K.one_hot(a, self.action_size)
        q_value = K.sum(py_x * a_one_hot, axis=1)
        error = K.abs(y - q_value)

        quadratic_part = K.clip(error, 0.0, 1.0)
        linear_part = error - quadratic_part
        loss = K.mean(0.5 * K.square(quadratic_part) + linear_part)

        optimizer = RMSprop(lr=0.00025, epsilon=0.01)
        updates = optimizer.get_updates(self.model.trainable_weights, [], loss)
        train = K.function([self.model.input, a, y], [loss], updates=updates)

        return train

    # approximate Q function using Convolution Neural Network
    # state is input and Q Value of each action is output of network 

def rpn_loss_regr(num_anchors):
    def rpn_loss_regr_fixed_num(y_true, y_pred):
        if K.image_dim_ordering() == 'th':
            x = y_true[:, 4 * num_anchors:, :, :] - y_pred
            x_abs = K.abs(x)
            x_bool = K.less_equal(x_abs, 1.0)
            return lambda_rpn_regr * K.sum(
                y_true[:, :4 * num_anchors, :, :] * (x_bool * (0.5 * x * x) + (1 - x_bool) * (x_abs - 0.5))) / K.sum(epsilon + y_true[:, :4 * num_anchors, :, :])
        else:
            x = y_true[:, :, :, 4 * num_anchors:] - y_pred
            x_abs = K.abs(x)
            x_bool = K.cast(K.less_equal(x_abs, 1.0), tf.float32)

            return lambda_rpn_regr * K.sum(
                y_true[:, :, :, :4 * num_anchors] * (x_bool * (0.5 * x * x) + (1 - x_bool) * (x_abs - 0.5))) / K.sum(epsilon + y_true[:, :, :, :4 * num_anchors])

    return rpn_loss_regr_fixed_num 

def rpn_bbox_loss_graph(config, target_bbox, rpn_match, rpn_bbox):
    """Return the RPN bounding box loss graph.

    config: the model config object.
    target_bbox: [batch, max positive anchors, (dy, dx, log(dh), log(dw))].
        Uses 0 padding to fill in unsed bbox deltas.
    rpn_match: [batch, anchors, 1]. Anchor match type. 1=positive,
               -1=negative, 0=neutral anchor.
    rpn_bbox: [batch, anchors, (dy, dx, log(dh), log(dw))]
    """
    # Positive anchors contribute to the loss, but negative and
    # neutral anchors (match value of 0 or -1) don't.
    rpn_match = K.squeeze(rpn_match, -1)
    indices = tf.where(K.equal(rpn_match, 1))

    # Pick bbox deltas that contribute to the loss
    rpn_bbox = tf.gather_nd(rpn_bbox, indices)

    # Trim target bounding box deltas to the same length as rpn_bbox.
    batch_counts = K.sum(K.cast(K.equal(rpn_match, 1), tf.int32), axis=1)
    target_bbox = batch_pack_graph(target_bbox, batch_counts,
                                   config.IMAGES_PER_GPU)

    # TODO: use smooth_l1_loss() rather than reimplementing here
    #       to reduce code duplication
    diff = K.abs(target_bbox - rpn_bbox)
    less_than_one = K.cast(K.less(diff, 1.0), "float32")
    loss = (less_than_one * 0.5 * diff**2) + (1 - less_than_one) * (diff - 0.5)

    loss = K.switch(tf.size(loss) > 0, K.mean(loss), tf.constant(0.0))
    return loss 

def jaccard_distance(y_true, y_pred, smooth=100):
    intersection = K.sum(K.abs(y_true * y_pred), axis=-1)
    sum_ = K.sum(K.abs(y_true) + K.abs(y_pred), axis=-1)
    jac = (intersection + smooth) / (sum_ - intersection + smooth)

    return (1 - jac) * smooth


###############################################################################
## OPTIMIZATIONS ##
############################################################################### 

def smooth_l1_loss(y_true, y_pred):
    """Implements Smooth-L1 loss.
    y_true and y_pred are typically: [N, 4], but could be any shape.
    """
    diff = K.abs(y_true - y_pred)
    less_than_one = K.cast(K.less(diff, 1.0), "float32")
    loss = (less_than_one * 0.5 * diff**2) + (1 - less_than_one) * (diff - 0.5)
    return loss 

def smooth_l1(gt, pred):
    # https://qiita.com/GushiSnow/items/8c946208de0d6a4e31e7
    diff = K.abs(gt - pred)
    less_than_one = K.cast(K.less(diff, 1.0), "float32")
    # difffが1より小さい場合、less_than_one==1
    loss = (less_than_one * 0.5 * diff**2) + (1 - less_than_one) * (diff - 0.5)
    return loss 

def trim_zeros_graph(boxes, name=None):
    """Often boxes are represented with matrices of shape [N, 4] and
    are padded with zeros. This removes zero boxes.

    boxes: [N, 4] matrix of boxes.
    non_zeros: [N] a 1D boolean mask identifying the rows to keep
    """
    non_zeros = tf.cast(tf.reduce_sum(tf.abs(boxes), axis=1), tf.bool)
    boxes = tf.boolean_mask(boxes, non_zeros, name=name)
    return boxes, non_zeros 

def rpn_bbox_loss_graph(config, target_bbox, rpn_match, rpn_bbox):
    """Return the RPN bounding box loss graph.

    config: the model config object.
    target_bbox: [batch, max positive anchors, (dy, dx, log(dh), log(dw))].
        Uses 0 padding to fill in unsed bbox deltas.
    rpn_match: [batch, anchors, 1]. Anchor match type. 1=positive,
               -1=negative, 0=neutral anchor.
    rpn_bbox: [batch, anchors, (dy, dx, log(dh), log(dw))]
    """
    # Positive anchors contribute to the loss, but negative and
    # neutral anchors (match value of 0 or -1) don't.
    rpn_match = K.squeeze(rpn_match, -1)
    indices = tf.where(K.equal(rpn_match, 1))

    # Pick bbox deltas that contribute to the loss
    rpn_bbox = tf.gather_nd(rpn_bbox, indices)

    # Trim target bounding box deltas to the same length as rpn_bbox.
    batch_counts = K.sum(K.cast(K.equal(rpn_match, 1), tf.int32), axis=1)
    target_bbox = batch_pack_graph(target_bbox, batch_counts,
                                   config.IMAGES_PER_GPU)

    # TODO: use smooth_l1_loss() rather than reimplementing here
    #       to reduce code duplication
    diff = K.abs(target_bbox - rpn_bbox)
    less_than_one = K.cast(K.less(diff, 1.0), "float32")
    loss = (less_than_one * 0.5 * diff**2) + (1 - less_than_one) * (diff - 0.5)

    loss = K.switch(tf.size(loss) > 0, K.mean(loss), tf.constant(0.0))
    return loss 

def trim_zeros_graph(boxes, name='trim_zeros'):
    """Often boxes are represented with matrices of shape [N, 4] and
    are padded with zeros. This removes zero boxes.

    boxes: [N, 4] matrix of boxes.
    non_zeros: [N] a 1D boolean mask identifying the rows to keep
    """
    non_zeros = tf.cast(tf.reduce_sum(tf.abs(boxes), axis=1), tf.bool)
    boxes = tf.boolean_mask(boxes, non_zeros, name=name)
    return boxes, non_zeros 

def rpn_bbox_loss_graph(config, target_bbox, rpn_match, rpn_bbox):
    """Return the RPN bounding box loss graph.

    config: the model config object.
    target_bbox: [batch, max positive anchors, (dy, dx, log(dh), log(dw))].
        Uses 0 padding to fill in unsed bbox deltas.
    rpn_match: [batch, anchors, 1]. Anchor match type. 1=positive,
               -1=negative, 0=neutral anchor.
    rpn_bbox: [batch, anchors, (dy, dx, log(dh), log(dw))]
    """
    # Positive anchors contribute to the loss, but negative and
    # neutral anchors (match value of 0 or -1) don't.
    rpn_match = K.squeeze(rpn_match, -1)
    indices = tf.where(K.equal(rpn_match, 1))

    # Pick bbox deltas that contribute to the loss
    rpn_bbox = tf.gather_nd(rpn_bbox, indices)

    # Trim target bounding box deltas to the same length as rpn_bbox.
    batch_counts = K.sum(K.cast(K.equal(rpn_match, 1), tf.int32), axis=1)
    target_bbox = batch_pack_graph(target_bbox, batch_counts,
                                   config.IMAGES_PER_GPU)

    # TODO: use smooth_l1_loss() rather than reimplementing here
    #       to reduce code duplication
    diff = K.abs(target_bbox - rpn_bbox)
    less_than_one = K.cast(K.less(diff, 1.0), "float32")
    loss = (less_than_one * 0.5 * diff**2) + (1 - less_than_one) * (diff - 0.5)

    loss = K.switch(tf.size(loss) > 0, K.mean(loss), tf.constant(0.0))
    return loss 

def smooth_l1_loss(y_true, y_pred):
    """Implements Smooth-L1 loss.
    y_true and y_pred are typically: [N, 4], but could be any shape.
    """
    diff = K.abs(y_true - y_pred)
    less_than_one = K.cast(K.less(diff, 1.0), "float32")
    loss = (less_than_one * 0.5 * diff**2) + (1 - less_than_one) * (diff - 0.5)
    return loss 

def huber_loss(y_true, y_predict):
    err = y_true - y_predict

    cond = K.abs(err) < HUBER_LOSS_DELTA
    L2 = 0.5 * K.square(err)
    L1 = HUBER_LOSS_DELTA * (K.abs(err) - 0.5 * HUBER_LOSS_DELTA)
    loss = tf.where(cond, L2, L1)

    return K.mean(loss) 

def class_loss_regr(num_classes):
	def class_loss_regr_fixed_num(y_true, y_pred):
		x = y_true[:, :, 4*num_classes:] - y_pred
		x_abs = K.abs(x)
		x_bool = K.cast(K.less_equal(x_abs, 1.0), 'float32')
		return lambda_cls_regr * K.sum(y_true[:, :, :4*num_classes] * (x_bool * (0.5 * x * x) + (1 - x_bool) * (x_abs - 0.5))) / K.sum(epsilon + y_true[:, :, :4*num_classes])
	return class_loss_regr_fixed_num 

def call(self, x, mask=None):
        mask = K.not_equal(K.sum(K.abs(x), axis=2, keepdims=True), 0)
        n = K.sum(K.cast(mask, 'float32'), axis=1, keepdims=False)
        x_mean = K.sum(x, axis=1, keepdims=False) / (n + 1)
        return K.cast(x_mean, 'float32') 

def jaccard_distance(y_true, y_pred, smooth=100):
    intersection = K.sum(K.abs(y_true * y_pred), axis=-1)
    sum_ = K.sum(K.abs(y_true) + K.abs(y_pred), axis=-1)
    jac = (intersection + smooth) / (sum_ - intersection + smooth)

    return (1 - jac) * smooth


###############################################################################
## OPTIMIZATIONS ##
############################################################################### 

def class_mae(y_true, y_pred):
    return K.mean(
        K.abs(
            K.argmax(y_pred, axis=-1) - K.argmax(y_true, axis=-1)
        ),
        axis=-1
    ) 

def trim_zeros_graph(boxes, name=None):
    """Often boxes are represented with matricies of shape [N, 4] and
    are padded with zeros. This removes zero boxes.

    boxes: [N, 4] matrix of boxes.
    non_zeros: [N] a 1D boolean mask identifying the rows to keep
    """
    non_zeros = tf.cast(tf.reduce_sum(tf.abs(boxes), axis=1), tf.bool)
    boxes = tf.boolean_mask(boxes, non_zeros, name=name)
    return boxes, non_zeros 

def class_loss_regr(num_classes):
    def class_loss_regr_fixed_num(y_true, y_pred):
        x = y_true[:, :, 4*num_classes:] - y_pred
        x_abs = K.abs(x)
        x_bool = K.cast(K.less_equal(x_abs, 1.0), 'float32')
        return lambda_cls_regr * K.sum(y_true[:, :, :4*num_classes] * (x_bool * (0.5 * x * x) + (1 - x_bool) * (x_abs - 0.5))) / K.sum(epsilon + y_true[:, :, :4*num_classes])
    return class_loss_regr_fixed_num 

def trim_zeros_graph(boxes, name=None):
    """Often boxes are represented with matricies of shape [N, 4] and
    are padded with zeros. This removes zero boxes.

    boxes: [N, 4] matrix of boxes.
    non_zeros: [N] a 1D boolean mask identifying the rows to keep
    """
    non_zeros = tf.cast(tf.reduce_sum(tf.abs(boxes), axis=1), tf.bool)
    boxes = tf.boolean_mask(boxes, non_zeros, name=name)
    return boxes, non_zeros 

def rpn_bbox_loss_graph(config, target_bbox, rpn_match, rpn_bbox):
    """Return the RPN bounding box loss graph.

    config: the model config object.
    target_bbox: [batch, max positive anchors, (dy, dx, log(dh), log(dw))].
        Uses 0 padding to fill in unsed bbox deltas.
    rpn_match: [batch, anchors, 1]. Anchor match type. 1=positive,
               -1=negative, 0=neutral anchor.
    rpn_bbox: [batch, anchors, (dy, dx, log(dh), log(dw))]
    """
    # Positive anchors contribute to the loss, but negative and
    # neutral anchors (match value of 0 or -1) don't.
    rpn_match = K.squeeze(rpn_match, -1)
    indices = tf.where(K.equal(rpn_match, 1))

    # Pick bbox deltas that contribute to the loss
    rpn_bbox = tf.gather_nd(rpn_bbox, indices)

    # Trim target bounding box deltas to the same length as rpn_bbox.
    batch_counts = K.sum(K.cast(K.equal(rpn_match, 1), tf.int32), axis=1)
    target_bbox = batch_pack_graph(target_bbox, batch_counts,
                                   config.IMAGES_PER_GPU)

    # TODO: use smooth_l1_loss() rather than reimplementing here
    #       to reduce code duplication
    diff = K.abs(target_bbox - rpn_bbox)
    less_than_one = K.cast(K.less(diff, 1.0), "float32")
    loss = (less_than_one * 0.5 * diff**2) + (1 - less_than_one) * (diff - 0.5)

    loss = K.switch(tf.size(loss) > 0, K.mean(loss), tf.constant(0.0))
    return loss 

def smooth_l1_loss(y_true, y_pred):
    """Implements Smooth-L1 loss.
    y_true and y_pred are typicallly: [N, 4], but could be any shape.
    """
    diff = K.abs(y_true - y_pred)
    less_than_one = K.cast(K.less(diff, 1.0), "float32")
    loss = (less_than_one * 0.5 * diff**2) + (1 - less_than_one) * (diff - 0.5)
    return loss 

def nrmse_c(y_true, y_pred):
    return K.sqrt(K.mean(K.sum(K.square(y_true - y_pred))) / K.abs(K.identity(y_true))) 

def smape(y_true, y_pred):
        """
        Symmetric mean absolute percentage error
        """
        return 100*K.mean(K.abs(y_true -y_pred)/K.maximum((K.abs(y_true) + K.abs(y_pred)/2),
                                                          K.epsilon()))