You can deploy this model and start earning money today!

Junyi-Shen/
Junyi-UNet-latent-diffusion-flowers

Name: Junyi-UNet-latent-diffusion-flowers
Availability: InStock

Views

No views yet

Task 2a: Auto-encoder

Auto-encoder Architecture:

Encoder:

Layer Type	In Channels	Out Channels	Kernel	Stride	Padding	Output Size
Input	–	–	–	–	–	96×96×3
Conv1	3	16	3×3	2	1	48×48×16
Conv2	16	32	3×3	2	1	24×24×32
Conv3	32	64	3×3	2	1	12×12×64

Decoder:

Layer Type	In Channels	Out Channels	Kernel	Stride	Padding	Output Size
Transpose Conv1	64	32	3×3	2	1	24×24×32
Transpose Conv2	32	16	3×3	2	1	48×48×16
Transpose Conv3	16	3	3×3	2	1	96×96×3

Training:

Results:

UNet Architecture:

Layer Name	Layer Type	In Channels	Out Channels	Kernel	Stride	Output Size
Input	–	–	–	–	–	12×12×64
Down1_1	Conv	64	128	3×3	1	12×12×128
Down1_2	Conv	128	128	3×3	1	12×12×128
Down1_3	Conv	128	128	3×3	1	12×12×128
Pool1	MaxPool	–	–	2×2	2	6×6×128
Down2_1	Conv	128	256	3×3	1	6×6×256
Down2_2	Conv	256	256	3×3	1	6×6×256
Down2_3	Conv	256	256	3×3	1	6×6×256
Pool2	MaxPool	–	–	2×2	2	3×3×256
Bottleneck1	Conv	256	512	3×3	1	3×3×512
Bottleneck2	Conv	512	1024	3×3	1	3×3×1024
Bottleneck3	Conv	1024	512	3×3	1	3×3×512
Bottleneck4	Conv	512	256	3×3	1	3×3×256
Up2_1	Transpose Conv	256	256	3×3	2	6×6×256
Up2_2	Conv	256	256	3×3	1	6×6×256
Up2_3	Conv	256+256	128	3×3	1	6×6×128
Up1_1	Transpose Conv	128	128	3×3	2	12×12×128
Up1_2	Conv	128	128	3×3	1	12×12×128
Up1_3	Conv	128+128	64	3×3	1	12×12×64
Output	Conv	64	64	3×3	1	12×12×64

Training:

Results: