本文yolov3的代码是阅读的https://github.com/qqwweee/keras-yolo3
本文yolov4的代码是阅读的https://github.com/bubbliiiing/yolov4-keras
yolov3的loss计算公式为:
参考:
AI研习社 - 研习AI产学研新知,助力AI学术开发者成长。
【论文理解】yolov3损失函数_dlut_yan的博客-CSDN博客_yolov3损失函数
yolov3代码中loss计算的相关代码:
xy_loss = object_mask * box_loss_scale * K.binary_crossentropy(raw_true_xy, raw_pred[...,0:2], from_logits=True)
wh_loss = object_mask * box_loss_scale * 0.5 * K.square(raw_true_wh-raw_pred[...,2:4])
confidence_loss = object_mask * K.binary_crossentropy(object_mask, raw_pred[...,4:5], from_logits=True)+ \
(1-object_mask) * K.binary_crossentropy(object_mask, raw_pred[...,4:5], from_logits=True) * ignore_mask
class_loss = object_mask * K.binary_crossentropy(true_class_probs, raw_pred[...,5:], from_logits=True)
yolov4的loss计算公式为:
参考:YOLOv4损失函数_CinzWS的博客-CSDN博客_yolov4损失函数
yolov4代码中loss计算的相关代码:
ciou_loss = object_mask * box_loss_scale * (1 - ciou)
confidence_loss = object_mask * K.binary_crossentropy(object_mask, raw_pred[..., 4:5], from_logits=True) + \
(1 - object_mask) * K.binary_crossentropy(object_mask, raw_pred[..., 4:5],
from_logits=True) * ignore_mask
class_loss = object_mask * K.binary_crossentropy(true_class_probs, raw_pred[..., 5:], from_logits=True)
以yolov4为例,讲讲一些细节.