Region-based Image Annotation using Asymmetrical Support Vector Machine-based Multiple-Instance Learning

摘要: In region-based image annotation, keywords are usually associated with images instead of individual regions in the training data set. This poses a major challenge for any learning strategy. this paper, we formulate annotation as supervised problem under Multiple-Instance Learning (MIL) framework. We present novel Asymmetrical Support Vector Machine-based MIL algorithm (ASVM-MIL), which extends conventional Machine (SVM) to setting by introducing asymmetrical loss functions false positives and negatives. The proposed ASVM-MIL is evaluated on both sets benchmark MUSK sets.