配置

1. 文件树配置
2. 环境配置
   1. conda env create -f environment.yaml
   2. pip install -r requirnments.txt

TODO：

• Run 1 KNN

  • 当前 Acc: 25% n=5
  • 交叉验证Acc 25%, cv=5, train_size=0.9
  • 注释
  • 添加preprocess
  • 代码结构美化
  • 添加PCA，KNN调参

• Run 2 Bag-of-visual-words

  • 当前 Acc: 0.06 (???)
  • Debug, 找出表现差的原因
  • 添加preprocess
  • 添加滑动窗口patches提取
  • 添加Img2Vector
  • 添加Img2Feature
  • 添加线形分类器集合

• Run 3 SIFT (冲点)

  • 使用先前的场景识别模型微调
  • [ ]

资源

https://www.cc.gatech.edu/classes/AY2016/cs4476_fall/results/proj4/html/euzun3/index.html

http://openimaj.org/tutorial/classification101.html

https://github.com/gary1346aa/Scene-Recognition

https://github.com/gurkandemir/Bag-of-Visual-Words

https://github.com/AMANVerma28/Indoor-Outdoor-scene-classification

https://github.com/buptchan/scene-classification

• Run3
  • https://blog.csdn.net/qq_36622009/article/details/102895411
  • https://blog.csdn.net/cliukai/article/details/102525486
  • https://blog.csdn.net/weixin_45666660/article/details/109104700
  • （代码示例）https://www.kaggle.com/code/pierre54/bag-of-words-model-with-sift-descriptors/notebook
  • (优化相关)https://github.com/shackenberg/phow_caltech101.py
  • （优化相关, ORB算法）https://zhuanlan.zhihu.com/p/261966288
  • https://github.com/beyzacevik/Scene-Recognition-using-SIFT