Weakly supervised object localization (WSOL) is a challenging problem which\naims to localize objects with only image-level labels. Due to the lack of\nground truth bounding boxes, class labels are mainly employed to train the\nmodel. This model generates a class activation map (CAM) which activates the\nmost discriminate features. However, the main drawback of CAM is the ability to\ndetect just a part of the object. To solve this problem, some researchers have\nremoved parts from the detected object \\cite{b1, b2, b4}, or the image\n\\cite{b3}. The aim of removing parts from image or detected parts of the object\nis to force the model to detect the other features. However, these methods\nrequire one or many hyper-parameters to erase the appropriate pixels on the\nimage, which could involve a loss of information. In contrast, this paper\nproposes a Hierarchical Complementary Learning Network method (HCLNet) that\nhelps the CNN to perform better classification and localization of objects on\nthe images. HCLNet uses a complementary map to force the network to detect the\nother parts of the object. Unlike previous works, this method does not need any\nextras hyper-parameters to generate different CAMs, as well as does not\nintroduce a big loss of information. In order to fuse these different maps, two\ndifferent fusion strategies known as the addition strategy and the l1-norm\nstrategy have been used. These strategies allowed to detect the whole object\nwhile excluding the background. Extensive experiments show that HCLNet obtains\nbetter performance than state-of-the-art methods.\n