def f1_score(y_true, y_pred):
"""calcultes f1 score """
true_positives = K.sum(K.round(K.clip(y_true * y_pred, 0, 1)))
possible_positives = K.sum(K.round(K.clip(y_true, 0, 1)))
predicted_positives = K.sum(K.round(K.clip(y_pred, 0, 1)))
precision = true_positives / (predicted_positives + K.epsilon())
recall = true_positives / (possible_positives + K.epsilon())
return 2 * ((precision * recall) / (precision + recall + K.epsilon()))
def f1_loss(y_true, y_pred):
tp = K.sum(K.cast(y_true * y_pred, 'float'), axis=0)
tn = K.sum(K.cast((1 - y_true) * (1 - y_pred), 'float'), axis=0)
fp = K.sum(K.cast((1 - y_true) * y_pred, 'float'), axis=0)
fn = K.sum(K.cast(y_true * (1 - y_pred), 'float'), axis=0)
p = tp / (tp + fp + K.epsilon())
r = tp / (tp + fn + K.epsilon())
f1 = 2 * p * r / (p + r + K.epsilon())
f1 = tf.where(tf.math.is_nan(f1), tf.zeros_like(f1), f1)
return 1 - K.mean(f1)
def R(y_true, y_pred):
true_positives = K.sum(
K.cast(K.greater(K.clip(y_true * y_pred, 0, 1), 0.20), 'float32'))
poss_positives = K.sum(
K.cast(K.greater(K.clip(y_true, 0, 1), 0.20), 'float32'))
recall = true_positives / (poss_positives + K.epsilon())
return recall
def P(y_true, y_pred):
true_positives = K.sum(
K.cast(K.greater(K.clip(y_true * y_pred, 0, 1), 0.20), 'float32'))
pred_positives = K.sum(
K.cast(K.greater(K.clip(y_pred, 0, 1), 0.20), 'float32'))
precision = true_positives / (pred_positives + K.epsilon())
return precision
def bbox_ciou (boxes1, boxes2):
boxes1_x0y0x1y1 = tf.concat([boxes1[..., :2] - boxes1[..., 2:] * 0.5,
boxes1[..., :2] + boxes1[..., 2:] * 0.5], axis=-1)
boxes2_x0y0x1y1 = tf.concat([boxes2[..., :2] - boxes2[..., 2:] * 0.5,
boxes2[..., :2] + boxes2[..., 2:] * 0.5], axis=-1)
boxes1_x0y0x1y1 = tf.concat([tf.minimum(boxes1_x0y0x1y1[..., :2], boxes1_x0y0x1y1[..., 2:]),
tf.maximum(boxes1_x0y0x1y1[..., :2], boxes1_x0y0x1y1[..., 2:])], axis=-1)
boxes2_x0y0x1y1 = tf.concat([tf.minimum(boxes2_x0y0x1y1[..., :2], boxes2_x0y0x1y1[..., 2:]),
tf.maximum(boxes2_x0y0x1y1[..., :2], boxes2_x0y0x1y1[..., 2:])], axis=-1)
boxes1_area = (boxes1_x0y0x1y1[..., 2] - boxes1_x0y0x1y1[..., 0]) * (
boxes1_x0y0x1y1[..., 3] - boxes1_x0y0x1y1[..., 1])
boxes2_area = (boxes2_x0y0x1y1[..., 2] - boxes2_x0y0x1y1[..., 0]) * (
boxes2_x0y0x1y1[..., 3] - boxes2_x0y0x1y1[..., 1])
left_up = tf.maximum(boxes1_x0y0x1y1[..., :2], boxes2_x0y0x1y1[..., :2])
right_down = tf.minimum(boxes1_x0y0x1y1[..., 2:], boxes2_x0y0x1y1[..., 2:])
inter_section = tf.maximum(right_down - left_up, 0.0)
inter_area = inter_section[..., 0] * inter_section[..., 1]
union_area = boxes1_area + boxes2_area - inter_area
iou = inter_area / (union_area + K.epsilon())
enclose_left_up = tf.minimum(boxes1_x0y0x1y1[..., :2], boxes2_x0y0x1y1[..., :2])
enclose_right_down = tf.maximum(boxes1_x0y0x1y1[..., 2:], boxes2_x0y0x1y1[..., 2:])
enclose_wh = enclose_right_down - enclose_left_up
enclose_c2 = K.pow(enclose_wh[..., 0], 2) + K.pow(enclose_wh[..., 1], 2)
p2 = K.pow(boxes1[..., 0] - boxes2[..., 0], 2) + K.pow(boxes1[..., 1] - boxes2[..., 1], 2)
atan1 = tf.atan(boxes1[..., 2] / (boxes1[..., 3] + K.epsilon()))
atan2 = tf.atan(boxes2[..., 2] / (boxes2[..., 3] + K.epsilon()))
v = 4.0 * K.pow(atan1 - atan2, 2) / (math.pi ** 2)
a = v / (1 - iou + v)
ciou = iou - 1.0 * p2 / enclose_c2 - 1.0 * a * v
return ciou
def FocalLoss(y_true, y_pred):
"""
:param y_true: A tensor of the same shape as `y_pred`
:param y_pred: A tensor resulting from a sigmoid
:return: Output tensor.
"""
gamma = 2.0
alpha = 0.25
pt_1 = tf.where(tf.equal(y_true, 1), y_pred, tf.ones_like(y_pred))
pt_0 = tf.where(tf.equal(y_true, 0), y_pred, tf.zeros_like(y_pred))
epsilon = K.epsilon()
# clip to prevent NaN's and Inf's
pt_1 = K.clip(pt_1, epsilon, 1. - epsilon)
pt_0 = K.clip(pt_0, epsilon, 1. - epsilon)
return -K.mean(alpha * K.pow(1. - pt_1, gamma) * K.log(pt_1)) \
- K.mean((1 - alpha) * K.pow(pt_0, gamma) * K.log(1. - pt_0))
def euclidean_distance(x, y):
sum_sqrt = K.sum(K.square(x - y), axis=1, keepdims=True)
return K.sqrt(K.maximum(sum_sqrt, K.epsilon()))