Mobile networks and binary neural networks are the most commonly used techniques for modern deep learning models to perform a variety of tasks on embedded systems. In this paper, we develop a technique to identify an object considering the deep learning pre-trained model MobileNet for Single Shot Multi-Box Detector (SSD). This algorithm is used for real-time detection, and for webcam feed to detect the purpose webcam which detects the object in a video stream. Therefore, we use an object detection module that can detect what is in the video stream. In order to implement the module, we combine the MobileNet and the SSD framework for a fast and efficient deep learning-based method of object detection. The main purpose of our research is to elaborate the accuracy of an object detection method SSD and the importance of pre-trained deep learning model MobileNet. The experimental results show that the Average Precision (AP) of the algorithm to detect different classes as car, person and chair is 99.76%, 97.76% and 71.07%, respectively. This improves the accuracy of behavior detection at a processing speed which is required for the real-time detection and the requirements of daily monitoring indoor and outdoor.

Real-Time Object Detection Using Pre-Trained Deep Learning Models MobileNet-SSD

The visual-inertial integrated navigation system (VINS) has been extensively studied over the past decades to provide accurate and low-cost positioning solutions for autonomous systems. Satisfactory performance can be obtained in an ideal scenario with sufficient and static environment features. However, there are usually numerous dynamic objects in deep urban areas, and these moving objects can severely distort the feature-tracking process which is critical to the feature-based VINS. One well-known method that mitigates the effects of dynamic objects is to detect vehicles using deep neural networks and remove the features belonging to surrounding vehicles. However, excessive feature exclusion can severely distort the geometry of feature distribution, leading to limited visual measurements. Instead of directly eliminating the features from dynamic objects, this study proposes to adopt the visual measurement model based on the quality of feature tracking to improve the performance of the VINS. First, a self-tuning covariance estimation approach is proposed to model the uncertainty of each feature measurement by integrating two parts: (1) the geometry of feature distribution (GFD); (2) the quality of feature tracking. Second, an adaptive M-estimator is proposed to correct the measurement residual model to further mitigate the effects of outlier measurements, like the dynamic features. Different from the conventional M-estimator, the proposed method effectively alleviates the reliance on the excessive parameterization of the M-estimator. Experiments were conducted in typical urban areas of Hong Kong with numerous dynamic objects. The results show that the proposed method could effectively mitigate the effects of dynamic objects and improved accuracy of the VINS is obtained when compared with the conventional VINS method.

Robust Visual-Inertial Integrated Navigation System Aided by Online Sensor Model Adaption for Autonomous Ground Vehicles in Urban Areas

The main objective of this research work is to solve multiple object tracking problems in a given frame, wherein the proposed model intends to identify and track various objects. The problem has been solved in three stages viz. detecting, identifying, and tracking the object in a particular zone, i.e., but it is observed that something more could be done in this field, mostly the MOT-A score was not up to the mark; hence the proposed research work utilized Kalman filters for obtaining enhanced results and compared the obtained MOT-A metric with previous works, and the results were good. Object detection and recognition occur via the YOLO algorithm, which enables us to classify the objects into 80 classes. Then, Motion Prediction and feature generation occur in which an estimation model is created, and Kalman filters are used to model these states for capturing moving objects in the frame. Finally, tracking takes place with Kalman filters in the previous frame, and newly detected objects are placed in the current frame, after which an association is made for new detection. All this is done via the DeepSORT algorithm, which is essentially a Deep association metric with the SORT algorithm. Here, Kalman filters are used as they improved the accuracy of the proposed model and yielded better results. On the same lines, YOLO is used to perform object detection and recognition at the same time. It is also a detector, which by applying a single neural network, it can predict the bounding boxes and perform multi-class classification. This problem can have various applications, especially in traffic management. It can also prevent people from gathering during COVID times and raising an alert for all the authorities. Hence, this is a multidisciplinary approach wherein the work of object detection is being used in various fields like crowd assembling, Surveillance, Animal management in Zoo/Biodiversity parks as well as in the case of traffic systems as well. The real motivation behind this work was using the state-of-the-art technology to solve the modern-day problems. Manually human vigilance in large areas is a utopian task and especially when surveillance and security is big threat out there. Adding to it is the COVID Pandemic which has claimed millions of lives and yet vaccination is a still a dream yet to be realized. Hence, cluster identification problem is considered and also the ways to solve this challenge using the state-of-the-art technology. Once it has been achieved, it is initiated further to find more applications of this technology in various other domains and found out how it is applied in various use cases in case of crowd gathering at a single point and how this is a similar problem in case of animal gathering at a point and how forest rangers can solve the problem in a more efficient way.

Object tracking and counting in a zone using YOLOv4, DeepSORT and TensorFlow

Deep learning based torsional nystagmus detection for dizziness and vertigo diagnosis

A few years back, when the image processing hardware and software were created, it was limited, and most of the time, object detection would fail., But as with time, the advancement in technology has significantly changed the scenario. A lot of researchers worked in this field to carry out a solution through which they can detect objects in any field, especially in the robotic domain [1]. In today's world, with so much research in the field of deep learning, it is very easy to identify and detect any object using computer vision. This paper focuses on the various deep learning technologies and algorithms through which object detection can be done. A new and advanced deep learning technology known as salient object detection has been discussed. Also, the 3D object detection and the end-to-end approach for object detection are discussed. The existing methods of deep learning through which object detection can be done. The applications in which object detection is being used and the importance of object detection. It also reports; what the predecessors have done, what problems have been solved by them, how they solved these problems, the characteristics of the predecessors' methods and their future work.

/pdf/the-object-detection-perspective-and-obstacles-in-robotic-a-3txofprc.pdf

The Object Detection, Perspective and Obstacles In Robotic: A Review

The object detection and tracking is the important steps of computer vision algorithm. The robust object detection is the challenge due to variations in the scenes. Another biggest challenge is to track the object in the occlusion conditions. Hence in this approach, the moving objects detection using TensorFlow object detection API. Further the location of the detected object is pass to the object tracking algorithm. A novel CNN based object tracking algorithm is used for robust object detection. The proposed approach is able to detect the object in different illumination and occlusion. The proposed approach achieved the accuracy of 90.88% on self generated image sequences.

Prof.Supriya Mangale

Papers

Moving Object Detection and Tracking Using Convolutional Neural Networks