Skip to main content

Table 1 Architecture of the Places CNNs model and the dimension of the feature of each layer

From: Matching-range-constrained real-time loop closure detection with CNNs features

 

Convolutional

Fully connected

Layer

CONV1

POOL1

CONV2

POOL2

CONV3

CONV4

CONV5

POOL5

FC6

FC7

FC8

Dimension

290,400

69,984

186,624

43,264

64,896

64,896

43,264

9216

4096

4096

1000