# Eff+CBAM Survey 2019-2021
###### tags: `Survey`

## Multi-Classification Study of the Tuberculosis with 3D CBAM-ResNet and EfficientNet
`醫療` `CLEF 2021`
- Data : ImageCLEF 2021 Tuberculosis – CT report challenge
`provided as 3D dataset`
- class : 5
- train : 917
- test : 421
- Each CT-image can correspond to only one TB type at a time
:question:Multi-Classification?
- multi-classification task
- CBAM-Resnet and EfficientNet B5 as backbone
- CBAM use for each block of the Resnet
- augmentation
- 亮度
- 剪切
- 縮放
- 翻轉
- Results

    kappa -> Cohen’s Kappa 一種評斷標準

## C + EffxNet: A novel hybrid approach for COVID-19 diagnosis on CT images based on CBAM and EfficientNet
`醫療` `Chaos, Solitons and Fractals Nonlinear Science, and Nonequilibrium and Complex Phenomena`
- Data
- [The first dataset](https://github.com/UCSD-AI4H/COVID-CT)
- 2 classes
- 349 COVID-19 images
- 396 non-Covid images
- [The second dataset ](https://www.kaggle.com/plameneduardo/sarscov2-ctscan-dataset)
- 2 classes
- 1601 COVID-19 images
- 1626 non-Covid images
- 20% use for test
-
- 利用 EfficientNet + CBAM 組成 Hybrid model
- 利用 Hybrid model 提出 feature
- 將 feature 送入 Principial component analysis (PCA) `一種統計技術，目的在降維`
- 最後用 SVM 分類
- Results

## Efficient Supervision Net: Polyp Segmentation Using EfficientNet and Attention Unit
`醫療` `MediaEval’20`
- Data : Medico polyp challenge dataset
- 1000 segmented images
- randomly split 70% for training and 30% for validation

- Inspired by the [multilevel hyper vision Net](https://arxiv.org/abs/1908.03339) and had properties of an encoder, decoder structure with supervision layers.
- Encoder block : Efficientnet B4
- Decoder block : Dense block + Concurrent Spatial and Channel Attention (CSCA) block
- Augmentation
- HorizontalFlip
- VerticalFlip
- Blur (limit = 3)
- Rotate (-10, 10)
- Results

## Classification of Amanita Species Based on Bilinear Networks with Attention Mechanism
`農業` `Agriculture 2021`
- 菇類病蟲害辨識
- Constructed the Bilinear convolutional neural networks (B-CNN) with attention mechanism model
- Data : self-built Amanita dataset
- 3219 Amanita images
- train : 80%
- test : 20%
- class : 7
- Augmentation
- random rotation
- translation
- cutting

- Bilinear convolutional neural networks
- Efficientnet B4
- CBAM
- Results


---
https://scholar.google.com/scholar?q=efficientnet+cbam+&hl=zh-TW&as_sdt=0%2C5&as_ylo=2019&as_yhi=