Skip to main content

Table 1 Parameters of the time-dependent encoder-decoder convolutional neural network (Enc-Dec CNN)

From: Atmospheric correction of vegetation reflectance with simulation-trained deep learning for ground-based hyperspectral remote sensing

Layer

Number of filters

Size of each filter

Stride

Output size

Input

Radiance

–

–

–

450 × 1

Encoder

Conv1

64

50

2

225 × 64

 

Conv2

128

25

2

113 × 128

 

Conv3

256

5

2

57 × 256

 

Conv4

64

256

1

57 × 64

Fully connected layers

FC\(_{rad}\)

–

–

–

64 × 1

 

FC\(_{day}\)

–

–

–

64 × 1

 

FC\(_{time}\)

–

–

–

64 × 1

 

FC\(_1\)

–

–

–

64 × 1

 

FC\(_2\)

–

–

–

64 × 1

 

FC\(_3\)

–

–

–

64 × 1

Decoder

ConvTrans1

256

5

2

128 × 256

 

ConvTrans2

128

25

2

256 × 128

 

ConvTrans3

64

50

2

512 × 64

 

ConvTrans4

1

64

1

512 × 1

 

FC\(_{dec}\)

–

–

–

450 × 1

Output

\(\Delta\)reflectance

–

–

–

450 × 1