| 摘要 |
The invention relates to a process for identifying objects in an image, and in which a rough locating of objects in the image is carried out with a first neural network (RZ1), thus providing locating assumptions. Next, a second neural network (RZ2) carries out a precise locating of an object in each assumption. Finally, a third neural network (RZ3) performs an identification of each object whose position is known with precision. Applications: - locating of countenances (faces) - remote monitoring - televisual audience rating. <IMAGE> |
