Automated Freshwater Fish Species Classification using Deep CNN

J. Inst. Eng. India Ser. B https://doi.org/10.1007/s40031-023-00883-2 ORIGINAL CONTRIBUTION Automated Freshwater Fish Species Classification using Deep CNN Jayashree Deka1 · Shakuntala Laskar1 · Bikramaditya Baklial2 Received: 9 February 2022 / Accepted: 8 April 2023 © The Institution of Engineers (India) 2023 Abstract Freshwater fish is considered a poor man’s protein supplement as they are easily available in lakes, rivers, natural ponds, paddy fields, beels, and fisheries. There are various freshwater fish species that resemble each other, making it difficult to classify them by their external appearance. Manual fish species identification always needs expertise and so, is erroneous. Recently, computer vision along with deep learning plays a significant role in underwater species classification research where the number of species under investigation is always limited to a maximum of eight (8). In this article, we choose deep-learning architectures, AlexNet and Resnet-50, to classify 20 indigenous freshwater fish species from the North-Eastern parts of India. The two models are fine-tuned for training and validation of the collected fish data. The performance of these networks is evaluated based on overall accuracy, precision, and recall rate. This paper reports the best overall classification accuracy, precision, and recall rate of 100% at a learning rate of 0.001 by the Resnet-50 model on our own dataset and benchmark Fish-Pak dataset. Comprehensive empirical analysis has proved that with an increasing Weight and Bias learning rate, the validation loss incurred by the classifier also increases. Keywords Fish classification · Fish species recognition · Resnet-50 · Deep-learning * Jayashree Deka 1 Electrical and Electronics Engineering, Assam Don Bosco University, Guwahati, Assam 781017, India 2 Department of Zoology, Bahona College, Jorhat, Assam 785101, India Introduction The aquaculture of South Asian countries like India, Pakistan, Bangladesh, Sri Lanka, and Nepal contributes in volume 7.9% to Asian and 7.1% to world fish production [1]. The predominant fish groups in these countries are the carps and the catfishes [2]. Among freshwater fishes, the Indian major carps are most cultured [3], followed by Exotic Carps, Minor Carps, Catfish, and Trout. Apart from these, there are plenty of small indigenous fish species (i.e.,-mola, climbing perch, barbs, batá, etc.) that grow in rivers, lakes, ponds, beels, streams, wetlands, lowland areas, and paddy fields. These indigenous fish are rich in nutrients, often containing high levels of zinc, iron, and vitamin A [4]. As per FishBase, India reports a total of 1035 freshwater fish species, while it is 241 from Pakistan and 256 from Bangladesh. Out of 1035 freshwater species, 450 are Small Indigenous Fish Species (SIFS, body length maximum 26 cm). Almost, 216 SIFS are reported from Assam, the North-Eastern part of India [5]. These indigenous fish species in the polyculture system have the potential of enhancing the nutritional security of the poor which in turn can provide greater employment opportunities in these underdeveloped Asian countries. As the production and marketing of these indigenous fish species are at the local level, they are mostly invisible in government statistics. Also, manual identification of fish species needs information about the number of fins, fin location, scales, and lateral line, head shape, color, and texture of the body. However, a lack of appropriate taxonomic knowledge or user expertise can potentially cause unsought repercussions for fishery management. Manual fish identification based on morphology is often considered an erroneous, inaccurate, and inefficient task. This shortcoming can be eliminated by using computer 13 Vol.:(0123456789) J. Inst. Eng. India Ser. B vision techniques, as they have shown outstanding performance in the various image-based classification task. Diversification of fish species in aquaculture has become a popular research topic in India and Bangladesh. In poly-culture fisheries, it has been observed that the presence of certain fish species (including SIFS) may result in an increase or decrease in the growth of some other species (mainly carp) [6–10]. So, it is important to identify/classify rival species for early segregation to promote better growth. The study of fish diversification is necessary to evaluate selectivity, dietary overlap, and food competition among the cultured species. Therefore, it is crucial to correctly classify all the fish species reared in one environment to attain the targeted sustainable growth, and productivity. The correct classification will also help fish farmers with biomass estimation, disease identification, feeding planning, and cost estimation. Based on computer vision algorithms, much research has been conducted for fish species recognition using fish images and underwater videos. Most of the recognition algorithms are based on hand-crafted feature extraction techniques (color, texture, and shapes) from images until the evolution of deep learning [11–22]. As deep-learning gains popularity in animal and plant species recognition, it becomes a common choice for fish classification research too. Li et al. 2022 have studied how recent machine vision techniques have changed the outlook of feature-based fish species classification experiments [23]. As the underwater fish images do not possess much color and texture information due to poor contrast and varied illumination, so only shape information plays a significant role in underwater fish image classification models. It has been observed that the research on fish identification mostly deals with marine-water fish resources and only a handful of research has been executed on fresh-water fish species. Meanwhile, a few researchers utilized pre-trained CNN models for fresh-water fish image classification where a constant background was maintained for all images captured in a laboratory environment [24–27]. Although these works demonstrate satisfactory performance, further research into the effect of varied backgrounds on the performance of these deep-learning models is still required. Moreover, there is still scope for improvement in the accuracy of the small-indigenous fish classification problems [24, 28]. Indian water bodies carry varieties of fresh-water fish species, still, we are unable to track down any research article focusing related to these species. Here, in this work, we perform a deep-learning-based based experiment using MATLAB to classify images of 20 varieties of fresh-water fish including small indigenous species that are native to Assam, India. We deploy a deep residual 50-layer network, ResNet-50, and an Alex-Net architecture to build a robust model to work on different background images. A total of 13 Fig. 1  Fish sample images used 587 images of these species are captured under different background conditions under natural lighting. Figure 1 s (...truncated)