Abstract Attribute based object recognition performs object recognition using the semantic properties of the object. Unlike the existing approaches that treat attributes as a middle level representation and require to estimate the attributes during testing, we propose to incorporate the hidden attributes, which are the attributes used only during training to improve model learning and are not needed during testing. To achieve this goal, we develop two different approaches to incorporate hidden attributes. The fifirst approach utilizes hidden attributes as additional information to improve the object classififi- cation model. The second approach further exploits the semantic relationships between the objects and the hidden attributes. Experiments on benchmark data sets demonstrate that both approaches can effectively improve the learning of the object classififiers over the baseline models that do not use attributes, and their combination reaches the best performance. Experiments also show that the proposed approaches outperform both state of the art methods that use attributes as middle level representation and the approaches that learn the classi- fifiers with hidden information