Article Preview
Top1. Introduction
Located in the extreme northeast of Algeria, El Kala National Park (PNEK) was created in 1983 and classified as a biosphere reserve in 1990 (Boulmaiz et al., 2020). PNEK is considered the most important site for breeding waterbirds in Algeria and one of the most important in the Mediterranean for threatened or restricted species. In 2020, more than 36 different species, including the white-headed duck, Ferruginous duck, and Marbled teal, were identified (Mammeria et al., 2019).
Heavily forested (more than 69% of its area), the PNEK extends over a 40 km coastal strip and runs along the Tunisian border for 98 km. More than 120,000 people live in this territory. This human pressure on wildlife species such as birds, makes them very vulnerable. In addition to the census of individuals of each species, the recognition of bird species has great potential importance in the assessment and prediction of the surrounding life, which positively influence the development of a country's economy and industry.
Considerable research in the field of computer vision and machine learning has resulted in various papers proposing and comparing methods for automatic recognition of bird species (Kahl et al., 2019; Stowell et al., 2019; LeBien et al., 2020). Most of the work has used vocalizations (specifically song) as the main features for bird classification (Nasirahmadi et al., 2020; Kahl et al., 2021; Xie et al., 2019; Xie et al., 2022; Stowell et al., 2019), nevertheless there is a limited database of vocalizations of waterbirds that do not have a specific song, compared to passerines that are songbirds (Bermúdez-Cuamatzin et al., 2018).
The practice of giving one or more labels to an image based on its content is known as image classification. This is one of the typical supervised learning issues. This issue asks the system to learn from a collection of labeled practice images and then identify the label of a new image when it is presented. In the past, there has been a lot of interest in large-scale image classification in the fields of computer vision and machine learning (El-Saadawy et al., 2021).
A component of artificial intelligence, machine learning allows computers to learn and develop on their own, without having to be explicitly programmed. In order to make better judgments in the future, the learning process begins with the interpretation of data, such as examples, a prior model, or suggestions. The main goal is to let the computers learn on their own, without interference or support from humans, and to remedy any errors through this learning (Hussien et al., 2020; Ganesan et al., 2022; Ramadan et al., 2022; Khamis et al., 2022, 2021; Inbarani et al., 2020, 2018, 2014a,b,c, 2015a,b; Aziz et al., 2013a,b,c; Azar et al., 2022, 2020, 2017, 2012; Jothi et al., 2022, 2020, 2019a,b, 2017, 2013; Nasser et al., 2021; Fouad et al., 2021; Samanta et al., 2018, ElBedwehy et al., 2014; Mukherjee et al., 2014).