0_overview

0. curse of high-dimension

1. the historty of object detection

2. iccv2009-multiclass

1. Sharing invariances
   • object recognition is invariant to rotation, translation, scaling, lighting, …
   • typical case: cnn(convolutional layer and pooling layer)
2. Sharing transformations
   • Transformations are shared
     and can be learnt from other tasks
   • style transfer
3. Sharing in constellation models
4. Sharing patches

3. cvpr2007-part3

1. classficator-discriminative methods
   • nearest neighbor
   • cnn
   • svm+kernel
   • CRF
2. classficator-boosting
   • Cascade of classifiers 【级联分类器，能综合各个分类器的优势】

4. slides of ICCV 2005 by Li Fei-Fei 【非常好的tutorial，值得反复看】

1. obj det 的方法的最早分类

   • bag of words models
   • parts-based models
   • discriminative methods
   • three main issues:
     • Representation
       How to represent an object category
     • Learning
       How to form the classifier, given training data
     • Recognition
       How the classifier is to be used on novel data

2. Representation

   • Generative / discriminative / hybrid
   • Appearance only or location and appearance
   • Invariances:
     View point
     Illumination
     Occlusion
     Scale
     Deformation
     Clutter
     etc.
   • Part-based or global w/sub-window
   • Use set of features or each pixel in image

3. Learning:

   • Unclear how to model categories, so we learn what distinguishes them rather than manually specify the difference -- hence current interest in machine learning)
   • Methods of training: generative vs. discriminative
     (generativie 是用高斯混合模型去拟合两类obj的样本点；discriminative是尝试找一个分界面来划分两类obj样本点；假设这里是两类obj)
   • Level of supervision
     Manual segmentation; bounding box; image labels; noisy labels
   • Batch/incremental (on category and image level; user-feedback )
   • Training images:
     Issue of overfitting
     Negative images for discriminative methods Priors
   • Priors

4. Recognition

   • Scale / orientation range to search over
   • Speed

5. Bag-of-words models
   见ppt 【这个tutorial ppt非常赞！需要时可以拿来反复揣摩】

6. part-based models
   见PPT 【依旧赞！】

7. discriminative models
   见ppt 【赞！】

8. concurrent segmentation and recognition
   见ppt 【赞！】

参考：

[1]. http://people.csail.mit.edu/torralba/shortCourseRLOC/index.html
[2]. 【"patch" in an image】 https://www.quora.com/What-is-a-patch-in-image-processing