[Model] VGG16
Jeff: fast.ai lesson 1&2感觉没讲什么干货。
~/keras/keras.json配置后台theano or tensorflow.
~/./theanorc处理器配置cpu or gpu.
VGG16的预测
VGG16的训练
VGG16结构分析
- GoogleNet略胜一筹
- 结构还算简单。
小卷积核3*3:
- Stack of three 3x3 conv (stride 1) layers has same effective receptive field as one 7x7 conv layer. 【第二层等价于5*5,第三层等价于7*7】
- But deeper, more non-linearities.
- And fewer parameters: 3 * (32C2) vs.72C2 for C channels per layer.
VGG性价比不怎么高感觉。
感觉googLeNet更适合移动设备,结构有点小复杂,但高效。
Tensorflow VGG16 and VGG19
Ref: https://github.com/machrisaa/tensorflow-vgg
Ref: models/research/slim/nets/vgg.py ---- 需要学会使用官网代码!