A Comparative Study on Object Detection Using Retinanet
B Srikanth Reddy, A. Mallikarjuna Reddy, M H D S Sradda, T Mounika, S Mounika, K Meghana · 2022 IEEE 2nd Mysore Sub Section International Conference (MysuruCon) · 2022
Object recognition has been useful in a variety of situations. In this, we have to select the selected regions from the image and have to classify them using a convolutional neural network. By using CNN we have to predict every selected region. RetinaNet seems to be the best effective algorithm in Deep learning for objects detection. The second algorithm based on the regression YOLOv3 method comes under this category. We predict the classes and the objects will be highlighted with the anchor boxes when multiple objects will display in a single frame using a RetinaNet. Yolo V3 the algorithm is fast as compared to other classification algorithms like Yolo. In real-time our algorithm process 45 frames per second. By capturing the streaming video for a specific time the system will give accurate results when compared with the non moving images. Detecting the objects from the images by considering the preprocessing techniques, Feature Extraction, RetinaNet and Kalman filter for tracking the object positions.