HOG+SVM on CIFAR-10

Dataset

• CIFAR-10

Requirements

• python
• scikit-learn
• scikit-image
• cPickle
• wxPython
• opencv

Usage

1. From https://drive.google.com/drive/folders/1ycDcidhsywSPBmgZuKSP-NdtrdIATVyr,

• Download the model to put it in ./
• Download the batches to put them in ./cifar-10-batches-py/
• Download the images to put them in ./images/

2. Model training and saving:

python train_save_test.py

3. Demo presentation with wxPython:

python wx_show_test.py

Chinese Introduction

• 通过sklearn和skimage提取hog特征，并且用svm进行物体识别。
• 用wxpython做了一个简单的界面，选择一张图片进行预测，图片来自CIFAR-10，总共10个类：'飞机','汽车','鸟','猫','鹿','狗','蛙','马','船','货车'。
• 因为为了节省测试的时间，然后就事先把训练出来的svm模型变量通过cPickle这个包序列化存储在本地上，测试时直接读取模型变量model就好。