In this work, we present a practical approach to the problem of facial\nlandmark detection. The proposed method can deal with large shape and\nappearance variations under the rich shape deformation. To handle the shape\nvariations we equip our method with the aggregation of manipulated face images.\nThe proposed framework generates different manipulated faces using only one\ngiven face image. The approach utilizes the fact that small but carefully\ncrafted geometric manipulation in the input domain can fool deep face\nrecognition models. We propose three different approaches to generate\nmanipulated faces in which two of them perform the manipulations via\nadversarial attacks and the other one uses known transformations. Aggregating\nthe manipulated faces provides a more robust landmark detection approach which\nis able to capture more important deformations and variations of the face\nshapes. Our approach is demonstrated its superiority compared to the\nstate-of-the-art method on benchmark datasets AFLW, 300-W, and COFW.\n