A Recent Trend in Individual Counting Approach Using Deep Network

Abstract

In video surveillance scheme, counting individuals is regarded as a crucial task. Of all the individual counting techniques in existence, the regression technique can offer enhanced performance under overcrowded area. However, this technique is unable to specify the details of counting individual such that it fails in locating the individual. On contrary, the density map approach is very effective to overcome the counting problems in various situations such as heavy overlapping and low resolution. Nevertheless, this approach may break down in cases when only the heads of individuals appear in video scenes, and it is also restricted to the feature's types. The popular technique to obtain the pertinent information automatically is Convolutional Neural Network (CNN). However, the CNN based counting scheme is unable to sufficiently tackle three difficulties, namely, distributions of non-uniform density, changes of scale and variation of drastic scale. In this study, we cater a review on current counting techniques which are in correlation with deep net in different applications of crowded scene. The goal of this work is to specify the effectiveness of CNN applied on popular individuals counting approaches for attaining higher precision results.

Underwater Fish Detection and Counting Using Mask Regional Convolutional Neural Network

“…Hence, computer vision has become vital in place of many manual inspection systems. They were beginning from hyperspectral images until nano images such as a satellite in a study by Anahita et al [1], cell detection by Yazan et al [2], optical character recognition by Tarik et al [3], vehicle counting by Abbas et al [4], and rice diseases diagnosis by Abdullah et al [5]. Furthermore, machine learning enhancement that can reverse feature engineering, namely deep learning, has attracted many researchers to explore the computer vision research area in place of handcrafted feature engineering.…”

“…Another significant matter in handcrafted feature engineering is scale-invariant. Many remarkable feature engineering inventions can address scale-invariant with the employment of a non-linear function [1] as Wavelet, SIFT, BOW, SURF; however, such approaches are less tolerable or robust when dealing with low-contrast and high-contrast images. Again, their procedures entail an extensive, long processing time for the training model.…”

A Conceptual Artificial Neural Network Model in Warehouse Receiving Management

“…Several Artificial Neural Network models can be learned to design a conceptual model application in industrial warehouse management. The current state-of-the-art object detection system, Faster R-CNN hypothesises bounding boxes, resamples pixels or features for each box, and applies a high-quality classifier [5]. The datasets of the Faster R-CNN were all from PASCAL VOC, MS COCO and ILSVRC [1]- [3], [6], 7].…”

A Study of Generalization and Fitness Landscapes for Neuroevolution

“…In neuroevolution, an EA is used to evolve weights, topologies and/or hyper-parameters of artificial neural networks. In this study, we focus on the evolution of convolutional neural networks (CNNs), because they are one of the most popular deep neural network architectures with applications including computer vision [14], [15], gestures recognition [16] and activity recognition [17]. They have a vast amount of tunable parameters that are difficult to set, which makes them perfect for testing the capabilities of neuroevolution.…”