Object Detection Using Tensorflow and Its Methods

EasyChair Preprint 9253

Abstract

TensorFlow, an open-source program, is used in this study to implement object identification using several models. As we’re all aware, object detection is a group of interconnected computer vision tasks that entails recognizing various things in a photograph, clip, or live broadcast. Both image classification and object identification employ object detection to identify objects in an image using multiple bounding boxes to accurately identify numerous objects in a video or image. Here, certain pre-trained models are used to recognize objects to make predictions or detect objects by determining whether the model is predictive. The different models used are Faster Region-Based Convolutional Neural Network (Faster-RCNN) is used to predict multiple objects from a digital image and with accuracy and the obtained accuracy of the model is 94%, Single Shot Detection (SSD) is used to predict the different objects from a video with accuracy and this model predicts the objects with an accuracy of 99.87%, and You Only Look Once (yolov5) is used to predicting the objects using Realtime i.e., using a webcam with accuracy this model performs the accuracy of 93%. The different accuracies that are obtained from the model are Predictable and it is working fine.

Keyphrases: Faster R-CNN, Single Shot Detector (SSD), TensorFlow, You look only once (YOLO)

@booklet{EasyChair:9253,
  author    = {Vg Hemant and L Praveen and Gangula Surendar and Karri Siva Somi Reddy and Abhishek Santra and Navjot Kaur and Balwinder Kaur Dhaliwal},
  title     = {Object Detection Using Tensorflow and Its Methods},
  howpublished = {EasyChair Preprint 9253},
  year      = {EasyChair, 2022}}