首页| JavaScript| HTML/CSS| Matlab| PHP| Python| Java| C/C++/VC++| C#| ASP| 其他|
购买积分 购买会员 激活码充值

您现在的位置是:虫虫源码 > 其他 > object-detection-implicit-shape-model

object-detection-implicit-shape-model

  • 资源大小:123.05 MB
  • 上传时间:2021-06-30
  • 下载次数:0次
  • 浏览次数:1次
  • 资源积分:1积分
  • 标      签:

资 源 简 介

The algorithm constructs an non-parametric implicit representation of an object at training time by creating clusters of the object"s parts (image patches extracted around SIFT interest points), keeping a spatial distribution relative to the centre point of the object, resulting in a codebook. Generalized Hough Transform is performed to vote for centre of the object in continuous space, and the maxima is found. Mean-Shift should be used to find the maxima, but for this implementation, something else was made up (which works most of the time, but definitely needs to be improved). The result is a bounding box around the object(s) computed from the image patches that contributed to the centres at the maxima of the distribution. 75% accuracy rate has been achieved, working best on images of the same scale as the used for training, so I suspect there was some over-fitting on the CALTech dataset, as their frontal face images are all of similar scale. The codebooks

相 关 资 源

您 可 能 感 兴 趣 的

同 类 别 推 荐

VIP VIP