M2dNet

• 1 主干网络
• 2 预测
• 3 训练
• 4 评估

M2dNet用TUM模块进行多次特征提取，利用FFM模块进行特征融合，又添加注意力机制增强模型特征提取能力。

1 主干网络

1.1 C3, C4, C5 = VGG16(inputs).outputs[1:]  # 提取基本特征
2.1 base_feature = FFMv1(C4, C5, feature_size_1=256, feature_size_2=512)
2.2 feature_pyramid = _create_feature_pyramid(base_feature, stage=4) # 图像金字塔
2.3 outputs = SFAM(feature_pyramid,feature_pyramid_sizes)  # 注意力机制
3.1 classification = keras.layers.Conv2D(filters=num_classes * num_anchors,kernel_size=3,strides=1,padding='same')(feature) # 预测类别
3.2 regression = keras.layers.Conv2D(filters=num_anchors * 4,kernel_size=3,strides=1,padding='same')(feature)  # 预测偏移

2 预测

3 训练

4 评估

标签：kernel,pyramid,检测,feature,目标,num,M2dNet,size
来源： https://blog.csdn.net/qq_35732321/article/details/114366596