Scholarlink

Real-time object detectionObject detection is examined to be one of the hardest and complex technologies of computer visionComputer vision field. It detects objects from the live videos, classifies frames and identifies the detected frames from the input. In past few years many models have been designed which had few limitations like specific object detectionObject detection, slow processing, low level accuracy and many more. The present Single Shot Multibox DetectorSingle Stage Detector (SSD) (SSD) model overcomes all the limitations. The model is fast and generates accurate outputs with minimum loss in detection. TensorFlow object detectionObject detection API framework is built above SSDSingle Stage Detector (SSD) algorithm trained on COCO dataset. SSDSingle Stage Detector (SSD) model uses MobileNet as backbone to extract feature maps from the live input frames. CNNCNN (ConvNet) classifiers predict objects from the extracted feature maps with detected object class name and bounding box with class probability. The model has obtained an accuracy of 97.6% trained at 100 epochs when compared with other object detectionObject detection models such as YOLOYOLO and R-CNNCNN (ConvNet). SSDSingle Stage Detector (SSD) model is identified as good model with height accuracy and real-time speed in detecting even the smaller objects.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Real-Time Object Detection and Classification Using Deep Learning Models

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Piyush Kumar Pareek,
  • B. R. Shobha Rani,
  • S. Bharathi

摘要

Real-time object detectionObject detection is examined to be one of the hardest and complex technologies of computer visionComputer vision field. It detects objects from the live videos, classifies frames and identifies the detected frames from the input. In past few years many models have been designed which had few limitations like specific object detectionObject detection, slow processing, low level accuracy and many more. The present Single Shot Multibox DetectorSingle Stage Detector (SSD) (SSD) model overcomes all the limitations. The model is fast and generates accurate outputs with minimum loss in detection. TensorFlow object detectionObject detection API framework is built above SSDSingle Stage Detector (SSD) algorithm trained on COCO dataset. SSDSingle Stage Detector (SSD) model uses MobileNet as backbone to extract feature maps from the live input frames. CNNCNN (ConvNet) classifiers predict objects from the extracted feature maps with detected object class name and bounding box with class probability. The model has obtained an accuracy of 97.6% trained at 100 epochs when compared with other object detectionObject detection models such as YOLOYOLO and R-CNNCNN (ConvNet). SSDSingle Stage Detector (SSD) model is identified as good model with height accuracy and real-time speed in detecting even the smaller objects.