Wyniki wyszukiwania - BazTech

1

Deep Learning for Small and Tiny Object Detection: A Survey

Kos Aleksandra, Belter Dominik, Majek Karol

Pomiary Automatyka Robotyka

|

2023

|

R. 27, nr 3

85--94

EN

In recent years, thanks to the development of Deep Learning methods, there has been significant progress in object detection and other computer vision tasks. While generic object detection is becoming less of an issue for modern algorithms, with the Average Precision for medium and large objects in the COCO dataset approaching 70 and 80 percent, respectively, small object detection still remains an unsolved problem. Limited appearance information, blurring, and low signal-to-noise ratio cause state-of-the-art general detectors to fail when applied to small objects. Traditional feature extractors rely on downsampling, which can cause the smallest objects to disappear, and standard anchor assignment methods have proven to be less effective when used to detect low-pixel instances. In this work, we perform an exhaustive review of the literature related to small and tiny object detection. We aggregate the definitions of small and tiny objects, distinguish between small absolute and small relative sizes, and highlight their challenges. We comprehensively discuss datasets, metrics, and methods dedicated to small and tiny objects, and finally, we make a quantitative comparison on three publicly available datasets.

PL

W ostatnich latach, dzięki rozwojowi metod uczenia głębokiego, dokonano znacznego postępu w detekcji obiektów i innych zadaniach widzenia maszynowego. Mimo że ogólne wykrywanie obiektów staje się coraz mniej problematyczne dla nowoczesnych algorytmów, a średnia precyzja dla średnich i dużych instancji w zbiorze COCO zbliża się odpowiednio do 70 i 80 procent, wykrywanie małych obiektów pozostaje nierozwiązanym problemem. Ograniczone informacje o wyglądzie, rozmycia i niski stosunek sygnału do szumu powodują, że najnowocześniejsze detektory zawodzą, gdy są stosowane do małych obiektów. Tradycyjne ekstraktory cech opierają się na próbkowaniu w dół, które może powodować zanikanie najmniejszych obiektów, a standardowe metody przypisania kotwic są mniej skuteczne w wykrywaniu instancji o małej liczbie pikseli. W niniejszej pracy dokonujemy wyczerpującego przeglądu literatury dotyczącej wykrywania małych i bardzo małych obiektów. Przedstawiamy definicje, rozróżniamy małe wymiary bezwzględne i względne oraz podkreślamy związane z nimi wyzwania. Kompleksowo omawiamy zbiory danych, metryki i metody, a na koniec dokonujemy porównania ilościowego na trzech publicznie dostępnych zbiorach danych.

2

An efficient pedestrian attribute recognition system under challenging conditions

Nguyen Ha X., Hoang Dong N., Tran Tuan A., Dang Tuan M.

Machine Graphics and Vision

|

2023

|

Vol. 32, No. 2

3--18

EN

In this work, an efficient pedestrian attribute recognition system is introduced. The system is based on a novel processing pipeline that combines the best-performing attribute extraction model with an efficient attribute filtering algorithm using keypoints of human pose. The attribute extraction models are developed based on several state-of-the-art deep networks via transfer learning techniques, including ResNet50, Swin-transformer, and ConvNeXt. Pre-trained models of these networks are fine-tuned using the Ensemble Pedestrian Attribute Recognition (EPAR) dataset. Several optimization techniques, including the advanced optimizer Adam with Decoupled Weight Decay Regularization (AdamW), Random Erasing (RE), and weighted loss functions, are adopted to solve issues of data unbalancing or challenging conditions like partial and occluded bodies. Experimental evaluations are performed via EPAR that contains 26 993 images of 1477 person IDs, most of which are in challenging conditions. The results show that the ConvNeXt-v2-B outperforms other networks; mean accuracy (mA) reaches 85.57%, and other indices are also the highest. The addition of AdamW or RE can improve accuracy by 1-2%. The use of new loss functions can solve the issue of data unbalancing, in which the accuracy of data-less attributes improves by a maximum of 14% in the best case. Significantly, when the attribute filtering algorithm is applied, the results are dramatically improved, and mA reaches an excellent value of 94.85%. Utilizing the state-of-the-art attribute extraction model with optimization techniques on the large-scale and diverse dataset and attribute filtering has shown a good approach and thus has a high potential for practical applications.

3

Real-time face mask detection in mass gatherings to reduce Covid-19 spread

Soner Swapnil, Litoriya Ratnesh, Khatri Ravi, Hussain Ali Asgar, Pagrey Shreyas, Kushwaha Sunil Kumar

Journal of Automation Mobile Robotics and Intelligent Systems

|

2023

|

Vol. 17, No. 1

51--58

EN

The Covid 19 (coronavirus) pandemic has become one of the most lethal health crises worldwide. This virus gets transmitted from a person by respiratory droplets when they sneeze or when they speak. According to leading and well‐known scientists, wearing face masks and maintain‐ ing six feet of social distance are the most substantial protections to limit the virus’s spread. In the proposed model we have used the Convolutional Neural Network (CNN) algorithm of Deep Learning (DL) to ensure efficient real‐time mask detection. We have divided the system into two parts—1. Train Face Mask Detector 2. Apply Face Mask Detector—for better understanding. This is a real‐ time application that is used to discover or detect the person who is wearing a mask at the proper position or not, with the help of camera detection. The system has achieved an accuracy of 99% after being trained with the dataset, which contains around 1376 images of width and height 224×224 and also gives the alarm beep message after the detection of no mask or improper mask usage in a public place.

4

New model of photovoltaic system adapted by a digital MPPT control and radiation predictions using deep learning in Morocco agricultural sector

Zouhri Amal, El Mallahi Mostafa

Journal of Automation Mobile Robotics and Intelligent Systems

|

2023

|

Vol. 17, No. 2

74--84

EN

Solar energy is an essential factor in Moroccan sustainable development, especially in solar pumping in the agricultural sector. It is therefore difficult to dissociate the energy system of a society from its economic development and social development. Solar radiation prediction is useful in giving us a global overview on maintaining the integrity of solar systems. Access to database use makes this process more flexible. Solar forecasts can be generated using various available data sources. There are two major pillars of this data: the exploitation of historical solar radiation data, and the exploitation of other meteorological factors. On the other hand, the choice of data can have an impact on the choice of the model and the approach employed. In this paper we suggest an idea that aims to monitor in real time the situation of solar radiation in Morocco, using Long Short‐Term Memory for deep learning models compared with Artificial Neural Networks and Deep Neural Networks to predict the solar radiation with regard to solar pumping in the Moroccan agricultural sector.