Effect on performance of image source and texton system used.

SIFT is the Scale Invariant Feature Transform, amd oBIFs are oriented Basic Image Features – alternative methods for analyzing local image structure. To allow comparison between rows, all scores are correct rates for the identification task given 64 already-known appearances. Baseline performance is 50%; confidence intervals are not wider than ±2.5%.

SIFT即尺度不变特征变换（Scale Invariant Feature Transform），oBIFs为定向基本图像特征（oriented Basic Image Features），二者均为用于分析局部图像结构的替代方法。为便于行与行之间的横向对比，所有评分均为在给定64种已知图像外观的前提下，对应识别任务的正确识别率。本次实验的基线性能为50%，对应的置信区间宽度不超过±2.5%。