预、API介绍
mxnet.metric
from mxnet import metric
cls_metric = metric.Accuracy()
box_metric = metric.MAE()
cls_metric.update([cls_target], [class_preds.transpose((0,2,1))])
box_metric.update([box_target], [box_preds * box_mask])
cls_metric.get()
box_metric.get()
gluon.loss.Loss
class FocalLoss(gluon.loss.Loss):
def __init__(self, axis=-1, alpha=0.25, gamma=2, batch_axis=0, **kwargs):
super(FocalLoss, self).__init__(None, batch_axis, **kwargs)
self._axis = axis
self._alpha = alpha
self._gamma = gamma
def hybrid_forward(self, F, output, label):
# Here `F` can be either mx.nd or mx.sym
# 这里使用F取代在forward中显式的指定两者,方便使用
# 所以非hybrid无此参数
output = F.softmax(output)
pj = output.pick(label, axis=self._axis, keepdims=True)
loss = - self._alpha * ((1 - pj) ** self._gamma) * pj.log()
return loss.mean(axis=self._batch_axis, exclude=True)
mxnet.contrib.ndarray.MultiBoxTarget
def training_targets(anchors, class_preds, labels):
"""
得到的全部边框坐标
得到的全部边框各个类别得分
真实类别及对应边框坐标
"""
class_preds = class_preds.transpose(axes=(0,2,1))
return MultiBoxTarget(anchors, labels, class_preds)
# Output achors: (1, 5444, 4)
# Output class predictions: (1, 5444, 3)
# batch.label: (1, 1, 5)
out = training_targets(anchors, class_preds, batch.label[0][0:1])
mxnet.contrib.ndarray.MultiBoxDetection