cyruschiu / image-recognition

Image recognition methods from bag of words (BoW), Spatial Pyramid Matching (SPM), Sparse Codeing SPM (ScSPM) to convolutional neural networks (CNN)

Image-recognition

Image recognition methods from bag of words (BoW), Spatial Pyramid Matching (SPM), Sparse Codeing SPM (ScSPM) to convolutional neural networks (CNN) and CNN-SVM.

Author: CyrusChiu @ntu

Included methods

• Bag of words [1]
• Spatial Pyramid Matching [2]
• Sparse Coding SPM [3]
• Convolutional neural networks [4]
• CNN-SVM [5]

Requirements

Basic (BoW, SPM, ScSPM)

• Python 2.7
• NumPy
• SciPy
• Scikit-learn
• OpenCV 3.0.0 + opencv_contrib installation instructions

We use OpenCV here to load the image and extract SIFT descriptor only, you can use any image library if you want.

Optional (CNN, CNN-SVM)

• keras
• Caffe, pycaffe installation instructions

Demo

example.py training a SVM with SPM method on Caltech101 Dataset [6]

$python example.py --train Caltech101/DatasetFile.txt

Usage

An end-to-end script with training and testing is provided

• Dataset
  Dataset File is a text file with the format of label and path like:

  #train.txt
  label path/to/train1.jpg  
  label path/to/train2.jpg  
  label path/to/train3.jpg  
  

  #test.txt
  label path/to/test1.jpg
  label path/to/test2.jpg
  

• Training and evaluation on training set, predict on test set which is labeled

  $python yourMethod.py --train path/to/train.txt --test path/to/test.py
  

References

[1] CSURKA, Gabriella, et al. Visual categorization with bags of keypoints. In: Workshop on statistical learning in computer vision, ECCV. 2004. p. 1-2.

[2] LAZEBNIK, Svetlana; SCHMID, Cordelia; PONCE, Jean. Beyond bags of features: Spatial pyramid matching for recognizing natural scene categories. In: Computer Vision and Pattern Recognition, 2006 IEEE Computer Society Conference on. IEEE, 2006. p. 2169-2178.

[3] Jianchao Yang, Kai Yu, Yihong Gong, and Thomas Huang. Linear spatial pyramid matching using sparse coding for image classification. CVPR2009

[4] Krizhevsky, Alex, Ilya Sutskever, and Geoffrey E. Hinton. "Imagenet classification with deep convolutional neural networks." Advances in neural information processing systems. 2012.

[5] Girshick, Ross, et al. "Rich feature hierarchies for accurate object detection and semantic segmentation." Computer Vision and Pattern Recognition (CVPR), 2014 IEEE Conference on. IEEE, 2014. 2014.

[6] http://www.vision.caltech.edu/Image_Datasets/Caltech101/

[7] http://www.cs.toronto.edu/~kriz/cifar.html

image-recognition's Issues

bow.py running error

Hello ，when I run bow.py I got this error in line 25
x_train = [each for each in zip(x_train, y_train) if each[0] != None]

ValueError: The truth value of an array with more than one element is ambiguous.Use a.any() or a.all()

Because each[0] here has more than one element and if I use each[0].any() or each[0].all() ，In the condition that each is None, It will get error also because None doesn't have the attribute any() or all()

Could you please tell me how can I solve this problem?Thanks a lot