การจำแนกภาพความสัมพันธ์ของแอ็คชันด้วยวิธีการโครงข่ายประสาทเทียมแบบสังวัตนาการ-ซัพพอร์ตเวกเตอร์แมชชีน

บทคัดย่อ

Image retrieval is an active problem in the digital image processing field. A large number of new techniques and systems have researcher involved and attempted to improve the problems. Therefore, we survey the theoretical and empirical contributions in the current decade related to content base image retrieval, keyword annotation, and automatic image retrieval process. The retrieval process of such keyword based approaches is done by keyword searching model. The model is rather rudimentary and it does not specific enough for representing the actual image retrieval. This paper presents a new approach to represent the interactions between object and action. The interaction relationships are including implied-by, type-of and mutually exclusive. The approach is composed of four main phases: (1) Keyword Annotation (2) Define Relationships (3) Relationship Predictions (4) Measurement and Evaluation. We train and test our model on a large scale image dataset of relationship actions. The experimental results indicate that our proposed approach offers significant performance improvements in the classification of relationship actions with maximum success rate of 74.4% in Data Set II.