# Dog Breed Classifier 投影片
**code**
[mobilenet_v2_130_224 Model](https://github.com/AmineAgrane/dog_breed_classification)
[模型剪枝實作](https://hackmd.io/@computerVision/r1VVFF-H_#2-Resnet56)
[CNN](https://github.com/Tiwarim386/Dog_Breed_Classifier_CNN)
## 1. 動機
## 2. 介紹dataset
* show照片：10222張
* 種類：120種狗狗
* 資料集分布
## 3. model design
* model: Resnet50
* 加入attention module
[CBAM tensorflow](https://github.com/zhangkaifang/CBAM-TensorFlow2.0)
[Resnet50添加attention](https://blog.csdn.net/qq_38253797/article/details/124267280)
* **CBAM**[介紹](https://zhuanlan.zhihu.com/p/65529934)
* Data augumentation
* **GridMask** Data Augmentation [介紹](https://zhuanlan.zhihu.com/p/103992528)
* 添加額外品種至資料集內(台灣土狗等等)
* Model compression(7選1)
->後來選的 [Soft Filter Pruning for Accelerating Deep Convolutional Neural Networks](https://arxiv.org/abs/1808.06866)

[github](https://github.com/Eric-mingjie/rethinking-network-pruning)
[論文](https://arxiv.org/abs/1810.05270)
## 4. 發展
* line bot
[參考1](https://medium.com/@andy6804tw/%E4%BD%BF%E7%94%A8heroku%E9%83%A8%E7%BD%B2%E6%A9%9F%E5%99%A8%E5%AD%B8%E7%BF%92api-eca61ea711f1)
* 系統架構圖