人脸检测，多实例,取多个实例来提取特征，形成一个特征池，分类器仍然是用级联生成，检测效果好，准确率比较高。-Face detection, multi-instance, take multiple instances to extract features, to form a feature pool with cascading classifier is still generating, testing effective, relatively high accuracy.

自己实现的 particle filter 结合 multiple instance learning 的代码。必须安装intel ipp 6.0，可在intel 的官方下载30天试用版。测试视频可从miltracker的官方网站下载-particle filter with multiple instance tracking

MIL论文及源码，基于opencv1.0和IPP6.0。已调通，调试中可能出现的问题及解决方法在压缩包中。-MIL(Multiple Instance Learning)paper and source code.The code is based on opencv1.0 and IPP6.0.