# tongu

**Repository Path**: teethwang/tongu

## Basic Information

- **Project Name**: tongu
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: MIT
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2019-06-28
- **Last Updated**: 2020-12-19

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# Coarse-to-fine hierarchical classification、iSQRT、Class-Balanced Focal Loss 等；同时，团队也提出「后验概率重校准」技术，即通过先验知识对模型输出的后验概率进行校准
自研的 Brain++ AutoML 技术。内部采用了 One-shot 神经架构搜索的方法
# Tongue feature 
# DFL-CNN : a fine-grained classifier
This is a simple pytorch re-implementation of CVPR 2018 [Learning a Discriminative Filter Bank Within a CNN for Fine-Grained Recognition](https://arxiv.org/pdf/1611.09932.pdf).

### Introduction:
+ Use VGG16 as base Network.
+ Part FCs is replaced by Global Average Pooling to reduce parameters.
+ Every some epoches, ten best patches is visualized in **vis_result** directory
+ Update: ResNet-101 DFL-CNN and Multi-scale DFL-CNN need to be done.
+ Data: dataset/train,datatest/test,dirname is the classname, build a soft syslink to original pics. 


### Algorithms Introduction:

 
 
### Results and Visualization of ten boxes for discriminative patches:
Run: python main.py
Results is in model/模型迭代记录.xlsx


### Note:
1. Visualization of ten best boxes is saved in **vis_result/**
2. Weight(checkpoint.pth.tar, model_best.pth.tar) is in **weight/**.
3. Loss info is saved in **log/**, accuracy is in log_test.txt.