Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/cypw/CRU-Net
https://github.com/cypw/CRU-Net
Last synced: about 1 month ago
JSON representation
- Host: GitHub
- URL: https://github.com/cypw/CRU-Net
- Owner: cypw
- Created: 2017-03-15T15:24:37.000Z (over 7 years ago)
- Default Branch: master
- Last Pushed: 2017-04-14T00:12:15.000Z (over 7 years ago)
- Last Synced: 2024-08-01T22:50:05.678Z (4 months ago)
- Size: 1.95 KB
- Stars: 45
- Watchers: 6
- Forks: 12
- Open Issues: 5
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
- awesome-image-classification - official : https://github.com/cypw/CRU-Net
- Awesome-MXNet - CRU-Net
- awesome-image-classification - official : https://github.com/cypw/CRU-Net
README
# Collective Residual Networks
This repository contains the code and trained models of "Sharing Residual Units Through Collective Tensor Factorization in Deep Neural Networks".## Implementation
### Augmentation
| Method | Settings |
| :------------- | :--------: |
| Random Mirror | True |
| Random Crop | 8% - 100% |
| Aspect Ratio | 3/4 - 4/3 |
| Random HSL | [20,40,50] |> Note:
> We did not use PCA Lighting and any other advanced augmentation methods.### Normalization
The augmented input images are substrated by mean RGB = [ 124, 117, 104 ], and then multiplied by 0.0167.
## Results
### ImageNet-1k
**Single crop validation error (center 224x224 crop from resized image with shorter side=256):**
Model|Setting|Model Size|Top-1
:-----|------:|---------:|:---:
CRU-Net-56 @x14|32x4d|98MB|21.9%
CRU-Net-56 @x14|136x1d|98MB|21.7%
CRU-Net-116 @x28x14|32x4d|168MB|20.6%
CRU-Net-116, wider @x28x14|64x4d|318MB|20.3%We also trained a tiny CRU-Net-56 with less than half the size of ResNet-50.
**Single crop validation error (center 224x224 crop from resized image with shorter side=256):**
Model|Setting|Model Size|Top-1
:-----|------:|---------:|:---:
CRU-Net-56,tiny @x14|32x4d|48MB|22.9%### Places365-Standard
**10-crop validation accuracy (averaging softmax scores of 10 224x224 crops from resized image with shorter side=256):**
Model|Setting|Model Size|Top-1
:----|------:|---------:|:---:
CRU-Net-116 @x28x14|32x4d|163MB|56.6%## Trained Models
Model|Setting|Dataset| Link
:----|------:|:------:|:---:
CRU-Net-56,tiny @x14|32x4d|ImageNet-1k|[GoogleDrive](https://goo.gl/oTG7HJ)
CRU-Net-56 @x14|32x4d|ImageNet-1k|[GoogleDrive](https://goo.gl/AD8Bs0)
CRU-Net-56 @x14|136x1d|ImageNet-1k|[GoogleDrive](https://goo.gl/PEY8al)
CRU-Net-116 @x28x14|32x4d|ImageNet-1k|[GoogleDrive](https://goo.gl/OFyIaD)
CRU-Net-116 @x28x14|32x4d|Places365-Standard|[GoogleDrive](https://goo.gl/6gglTz)