This paper compares three top methods for finding and identifying objects in images: Faster R-CNN, Mask R-CNN, and EfficientDet. The study tested these methods on a large set of images called the Microsoft COCO dataset, specifically focusing on how well they work when used to analyze videos in real-time. This research tackles the important problem of making object detection systems both very accurate and fast enough for use in things like self-driving cars, security systems, and robotics. The methodology includes data preprocessing, which involves frame extraction, resizing, and augmentation, along with model training using optimized hyperparameters learning rate scheduling and batch size tuning. Performance evaluation is carried out using metrics such as accuracy, precision, recall, F1 score, and total loss. The results indicate that EfficientDet achieves the highest accuracy of 90% and the lowest total loss of 0.28, outperforming Faster R-CNN and Mask R-CNN across key metrics. This study offers practical insights for selecting object detection models suited to real-time applications and addresses challenges such as occlusion, varying object sizes, and detection speed.