In this paper, we propose a method for 3D-model retrieval from one or more photos. This method provides an "optimal" selection of 2D views to represent a 3D-model, and a probabilistic Bayesian method for 3D-model retrieval from realistic photos and sketches using these views. The characteristic view selection algorithm is based on an adaptive clustering algorithm and uses statistical model distribution scores to select the optimal number of views. We also introduce a Bayesian approach to score the probability of correspondence between the queries and the 3D-models. We present our results on the Princeton 3D Shape Benchmark database (1814 3D-models) and 50 photos (real photographs, sketches, synthesised images). A practical on-line 3D-model retrieval system based on our approach is available on the web to asset our results [1].