We present an approach to multi-objective exploration whose goal is to autonomously explore an unknown indoor environment. Our objective is to build a semantic map containing highlevel information, namely rooms and the objects laid in these rooms. This approach was developed for the Panoramic and Active Camera for Object Mapping (PACOM) project in order to participate in a French exploration and mapping contest called CAROTTE2. To achieve efficient exploration, we combine two classical approaches: frontier-based exploration for 2D laser metric mapping and nextbest view computation for visual object search. Based on a stochastic sampling strategy, this approach looks for a position that maximizes a multi-objective cost function. We show the advantage of using this combined approach compared to each particular approach in isolation. Additionally, we show how an uncertainty reduction strategy makes it possible to reduce object localization error after exploration.