Detection and recognition of object is challenging task nowadays in the computer vision domain. To solve this problem, many objects detection technique introduced but it is very difficult for automation of object detection, in natural environment. In forest areas, Wild animal and vehicle detection is a dynamic research field since the last several decades. Novel DP-RCNN architecture has proposed in this paper, by combining detecting and classifying task in a single neural network. It is based on generating proposals by SPP layer and features extraction from proposals to train the network which provide high classified output. The network was trained by animal and vehicle dataset which are selected from COCO dataset with their labels respectively for different animal classes. Based on these stored datasets, the video frame has processed and classified using DP-RCNN method. This experiment is conducted to access the following qualities, say, to verify whether the system is able to detect the object and to measure the accuracy of the proposed system. Implementation result shows that proposed DP-RCNN improves accuracy and recall by 94.59 %, 0.97. Precision and recognition rate of proposed method also achieved by 0.95, 73.4% when compared with existing scheme.