Learning Condition-Invariant Scene Representations for Place Recognition across the Seasons Using Auto-Encoder and ICA

Table 5

Summary of the CAE (encoder and decoder) networks.

	Size of filters	Number of filters
Layers (encoder)
Convolutional layer 1	11 × 11	96
Max-pooling layer 1	2 × 2	96
Convolutional layer 2	5 × 5	256
Convolutional layer 3	3 × 3	384
Max-pooling layer 2	2 × 2	384
Layers (decoder)
Up-pooling layer 1	2 × 2	384
Convolutional layer 1	3 × 3	384
Convolutional layer 2	5 × 5	256
Up-pooling layer 2	2 × 2	96
Convolutional layer 3	11 × 11	96