The deep learning models enhanced object detection in videos to a very large scale. Object detection is a prime task in self-driven cars, satellite images, robotics, etc. The researchers are working to improve the efficiency of deep learning models for better object detection. The object detection models broad categories into one-stage and two-stage detectors. The current work focused on improvement in accuracy and speed of one stage detector with the help of hyper-parameter tuning. The earlier researcher has shown that YOLO and R-CNN are the appropriate models for real-time object detection. In this paper, a custom CNN model is given with hyper-parameter tuning and the results are compared with regions with convolutional neural networks, Fast regions with convolutional neural networks, Faster Region with CNN, and YOLO. The impact of hyper-parameter tuning on the result of CNN models for object detection is shown in this paper. The results are verified on live video dataset.