Ensemble of Exemplar-SVMs for Object Detection and Beyond(2011)学习笔记
2017-07-26 12:14
543 查看
这篇文章提出一种思想:不将同一大类物体放在一起训练,而将视觉上类似的一个(类)物体作为exemplar(范式)正样本分别训练成若干SVM,使用多个被校准的SVM进行物体的检测。
每个exemplar
E使用HOG描述,记为xE:8个pixel对应1个cell,共100cell;负样本NE维数相同。每个Exemplar-SVM
(wE; bE)用以区分xE和NE。
由于exemplar-SVM是分别单独训练的,校准(calibration)对于同一类的不同exemplar-SVM是重要的。Therefore,
we let each exemplar select its ownpositives and then use the SVM output scores on these positives, in addition tolots of held-out negatives, to calibrate the Exemplar-SVM.
在验证数据集中,对于不同exemplar-SVM检测结果与真值比较。检测结果与真值重合达50%以上为positive,小于20%为negative。校准过程实际是决策边界的调整,表现差的exemplar被抑制,决策线向exemplar移动;反之亦然。校准使得性能得到了提升。
每个exemplar
E使用HOG描述,记为xE:8个pixel对应1个cell,共100cell;负样本NE维数相同。每个Exemplar-SVM
(wE; bE)用以区分xE和NE。
由于exemplar-SVM是分别单独训练的,校准(calibration)对于同一类的不同exemplar-SVM是重要的。Therefore,
we let each exemplar select its ownpositives and then use the SVM output scores on these positives, in addition tolots of held-out negatives, to calibrate the Exemplar-SVM.
在验证数据集中,对于不同exemplar-SVM检测结果与真值比较。检测结果与真值重合达50%以上为positive,小于20%为negative。校准过程实际是决策边界的调整,表现差的exemplar被抑制,决策线向exemplar移动;反之亦然。校准使得性能得到了提升。
相关文章推荐
- Exemplar-SVMs for Object Detection and Beyond—Overview(一)
- Exemplar-SVMs for Object Detection and Beyond—Overview(一)
- Exemplar-SVMs for Object Detection and Beyond--Overview(一)
- Exemplar-SVMs for Object Detection and Beyond--Details (二)
- Exemplar-SVMs for Object Detection and Beyond—Details(二)
- Exemplar-SVMs for Object Detection and Beyond--Details (二)
- R-CNN学习笔记2:Rich feature hierarchies for accurate object detection and semantic segmentation
- 深度学习论文笔记:Rich feature hierarchies for accurate object detection and semantic segmentation
- [深度学习论文笔记][Object Detection] Rich feature hierarchies for accurate object detection and semantic seg
- RCNN系列学习笔记(1):Rich feature hierarchies for accurate object detection and semantic segmentation
- rcnn学习笔记:Rich feature hierarchies for accurate object detection and semantic segmentation
- 【转】R-CNN学习笔记2:Rich feature hierarchies for accurate object detection and semantic segmentation
- 《Rich feature hierarchies for accurate object detection and semantic segmentation》笔记
- 【深度学习:目标检测】RCNN学习笔记(2):Rich feature hierarchies for accurate object detection and semantic segmentat
- Rich featureHierarchies for accurate object detection and semantic segmentation 阅读笔记
- [深度学习论文笔记][CVPR 17 oral]A-Fast-RCNN: Hard Positive Generation via Adversary for Object Detection
- 论文阅读笔记:R-CNN:Rich feature hierarchies for accurate object detection and semantic segmentation
- Rich feature hierarchies for accurate object detection and semantic segmentation论文笔记
- 学习笔记 -- Win32多线程程序设计(二) WaitForSingleObject and WaitForMultipleObjects
- [深度学习论文笔记][Instance Segmentation] Hypercolumns for Object Segmentation and Fine-Grained Localization