Object Detection

Object Detection has been one of the areas of interest of research community for over years and has made significant advances in its journey so far. There is a tremendous scope in the applications that would benefit with more innovations in the domain of object detection. Rapid growth in the field of machine learning has complemented the efforts in this area and in the recent times, research community has contributed a lot in real time object detection. In the current work, authors have implemented real time object detection and have made efforts to improve the accuracy of the detection mechanism. In the current research, we have used ssd_v2_inception_coco model as Single Shot Detection models deliver significantly better results. A dataset of more than 100 raw images is used for training and then xml files are generated using labellimg. Tensor flow records generated are passed through training pipelines using the proposed model. OpenCV captures real-time images and CNN performs convolution operations on images. The real time object detection delivers an accuracy of 92.7%, which is an improvement over some of the existing models already proposed earlier. Model detects hundreds of objects simultaneously. In the proposed model, accuracy of object detection significantly improvises over existing methodologies in practice. There is a substantial dataset to evaluate the accuracy of proposed model. The model may be readily useful for object detection applications including parking lots, human identification, and inventory management.

Day by day demand for object tracing is increasing because of the huge scope in real-time applications. Object tracing is one of the difficult issues in the computer vision and video processing field. Nowadays, object tracing is a common problem in many applications specifically video footage, traffic management, video indexing, machine learning, artificial intelligence, and many other related fields. In this paper, the Enhanced Method of Object Tracing Using Extended Kalman Filter via Binary Search Algorithm is proposed. Initially, the background subtraction method was used for merge sort and binary search algorithm to identify moving objects from the video. Merge sort is to divide the regions and conquer the algorithm that arranges the region in ascending order. After sorting, the binary search algorithm detects the position of noise in sorted frames and then the next step extended the Kalman Filter algorithm used to predict the moving object. The proposed methodology is linear about the valuation of mean and covariance parameters. Finally, the proposed work considered less time as compared to the state of art methods while tacking the moving objects. Its shows less absolute error and less object tracing error while evaluating the proposed work.

The aim of this study is to evaluate the performance of the pre-trained models and compare them with the probability percentage of prediction in terms of execution time. This study uses the COCO dataset to evaluate both pre-trained image recognition and object detection, models. The results revealed that Tiny-YoloV3 is considered the best method for real-time applications as it takes less time. Whereas ResNet 50 is required for those applications which require a high probability percentage of prediction, such as medical image classification. In general, the rate of probability varies from 75% to 90% for the large objects in ResNet 50. Whereas in Tiny-YoloV3, the rate varies from 35% to 80% for large objects, besides it extracts more objects, so the rise of execution time is sensible. Whereas small size and high percentage probability makes SqueezeNet suitable for portable applications, while reusing features makes DenseNet suitable for applications for object identification.