https://github.com/ffiirree/encoder-decoder
What features are "U" networks needed for image segmentation?
https://github.com/ffiirree/encoder-decoder
autoencoder pytorch unet
Last synced: about 2 months ago
JSON representation
What features are "U" networks needed for image segmentation?
- Host: GitHub
- URL: https://github.com/ffiirree/encoder-decoder
- Owner: ffiirree
- Created: 2021-03-25T02:55:03.000Z (about 4 years ago)
- Default Branch: master
- Last Pushed: 2021-04-06T12:31:22.000Z (about 4 years ago)
- Last Synced: 2025-02-13T07:41:43.484Z (3 months ago)
- Topics: autoencoder, pytorch, unet
- Language: Python
- Homepage:
- Size: 35.2 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# EncoderDecoder
What features are "U" / Encoder-Decoder networks needed for image segmentation?
## EncoderDecoder Networks
- [x] AutoEncoder
- [ ] AutoEncoder with shortcuts
- [x] AutoEncoder with two decoders
- [x] `baseline:` UNet
- [x] UNet with two decoders
- [x] UNet 2+ / (modified) UNet 3+
- [x] modified UNet 3+ with two decoders
## with Carvana dataset
Net / Dice | vanilla | image reconstruction | object reconstruction | border reconstruction
|--|:--:|:--:|:--:|:--:|
| AE |98.80
98.52
98.25
98.15
| AE - 2 decoders | - | 98.11
98.83
98.27
98.46 | 98.64
98.75
98.57
98.13
| AE with shortcuts |
| AE with shortcuts - 2 decoders | - |
| UNet(`baseline`) |98.71
98.62
98.87
98.56
| UNet - 2 decoders | - | 98.78
98.81
98.65
98.92 | 99.08
98.29
98.08
98.98
| UNet 2+ |
| UNet 2+ - 2 decoders | - |
| UNet 3+ | 98.79
99.10
98.82
98.84 |
| UNet 3+ - 2 decoders | - | 98.89
99.12
99.02
99.23 |98.84
98.74
98.83
99.13
## with LITS 2017 dataset
Net| vanilla | image reconstruction | object reconstruction | border reconstruction
|--|:--:|:--:|:--:|:--:|
| AE |
| AE - 2 decoders |
| AE with shortcuts |
| AE with shortcuts - 2 decoders |
| UNet(`baseline`) |
| UNet - 2 decoders |
| UNet 2+ |
| UNet 2+ - 2 decoders |
| UNet 3+ |
| UNet 3+ - 2 decoders |