<p>Starting from the use of deep belief networks as a fast learning algorithm, deep learning techniques have proved to be efficient in number of ways because of their capability of learning self-defined features rather than operating on fed features. The concept has drawn ever-increasing research interest in recent years and has been achieving remarkable performance in learning tasks involving text, sound, or image/video components. Due to its promising performance, there have been many efforts made to evolve deep learning to better suit to various applications that may gain substantial improvement over the state-of-the-art methods. With an increased demand of autonomous vehicles, smart video surveillance systems, facial gesture, human detection applications, etc., fast and accurate object detection systems constitute a very promising area of research and development. Robust object detection requires recognition and classification of every object in an image under unfamiliar environmental conditions such as occlusion and cluttered objects in a confined space. Therefore, localizing and labeling each object in an image by drawing an appropriate bounding box around it makes object detection a challenging task. The article reviews methods which rely on the application of deep learning to object detection and recognition. A cutting-edge review is accomplished on various deep learning architectures and frameworks, summarizing recent progress in the domain. The paper also discusses promising future research directions by elaborating on the limitations of the deep learning techniques applied for detection and recognition of objects.</p>