Search

Identifying and counting fish are crucial for managing stocking, harvesting, and marketing of farmed fish. Researchers have used convolutional networks for these tasks and explored various approaches to enhance network learning. Batch normalization is one technique that improves network stability and accuracy. This study aimed to evaluate machine learning for identifying and counting pirapitinga Piaractus brachypomus fry with different batch sizes. The researchers used one thousand photographic images of Pirapitinga fingerlings, labeled with bounding boxes. They trained the adapted convolutional network model with batch normalization layers added at the end of each convolution block. They set the training to one hundred and fifty epochs and tested batch sizes of 5, 10, and 20. Furthermore, they measured network performance using precision, recall, and mAP@0.5. Models with smaller batch sizes performed less effectively. The training with a batch size of 20 achieved the best performance, with a precision of 96.74%, recall of 95.48%, mAP@0.5 of 97.08%, and accuracy of 98%. This indicates that larger batch sizes improve accuracy in detecting and counting pirapitinga fry across different fish densities.

Aquaculture produces more than 122 million tons of fish globally. Among the several economically important species are the Serrasalmidae, which are valued for their nutritional and sensory characteristics. To meet the growing demand, there is a need for automation and accuracy of processes, at a lower cost. Convolutional neural networks (CNNs) are a viable alternative for automation, reducing human intervention, work time, errors, and production costs. Therefore, the objective of this work is to evaluate the efficacy of convolutional neural networks (CNNs) in counting round fish fingerlings (Serrasalmidae) at different densities using 390 color photographs in an illuminated environment. The photographs were submitted to two convolutional neural networks for object detection: one model was adapted from a pre-trained CNN and the other was an online platform based on AutoML. The metrics used for performance evaluation were precision (P), recall (R), accuracy (A), and F1-Score. In conclusion, convolutional neural networks (CNNs) are effective tools for detecting and counting fish. The pre-trained CNN demonstrated outstanding performance in identifying fish fingerlings, achieving accuracy, precision, and recall rates of 99% or higher, regardless of fish density. On the other hand, the AutoML exhibited reduced accuracy and recall rates as the number of fish increased.

Developing a rearing technology requires the determination of the optimum stocking density. This study aimed to determine the best stocking density for Brycon orthotaenia juveniles grown in net cages, during the rearing phase. The research was conducted at the Experimental Farm of EPAMIG, in Felixlândia, Minas Gerais State, in the Três Marias Reservoir. 9,000 fingerlings were distributed into 12 net cages, 2 m3each, according to a completely randomized design with three replications and four different stocking densities (150, 300, 450, 600 fish m-3). Random samples were taken from each net cage, at the onset of the experiment and at 60 rearing days, to assess production parameters, specific growth rate and uniformity. We verified a positive linear relationship (p < 0.05) for final biomass (r2 = 0.88), weight gain (r2 = 0.87), productivity (r2 = 0.86), apparent feed conversion (r2 = 0.96) and economic viability (r2 = 0.96). For survival (r2 = 0.98), the relationship was negative (p < 0.05). There were no significant differences between treatments (p > 0.05) for growth parameters, uniformity and final weight. It can be concluded that the density of 300 fish m-3 is the most suitable, because it provides higher net revenue, survival, and a good feed conversion.

O objetivo deste trabalho foi determinar a densidade de estocagem de matrinxã (Brycon orthotaenia) em tanques-rede, considerando o desempenho produtivo. Foram utilizados 5.280 juvenis de Brycon orthotaenia com peso inicial de 115,2g±5,7, os quais foram distribuídos em 16 tanques-rede de 4³m nas densidades de 45, 70, 95 e 120 peixes m-3 em quatro repetições. Os peixes foram alimentados com ração comercial de 40% de proteína bruta durante o período experimental. Os dados foram submetidos à análise de variância e à regressão linear para estimativa do melhor ajuste a 5% de probabilidade. Os parâmetros produtivos como sobrevivência, peso inicial, peso final, conversão alimentar e taxa de crescimento específico não foram influenciados pela densidade de estocagem, assim como os parâmetros de qualidade da água. Entretanto, para a variável biomassa final, ganho de biomassa e consumo de ração total, o efeito foi significativo e estas apresentaram relação linear positiva com a densidade de estocagem. Avaliando os resultados, pode-se concluir que a densidade de estocagem de matrinxã (Brycon orthotaenia) até 120 peixes m-3 é a que proporciona maior produtividade e não representa efeitos negativos sobre o desempenho dos peixes.The objective of this study was to determine the stocking density of matrinxã (Brycon orthotaenia) in cages, considering the productive performance. Juvenile Brycon orthotaenia (n = 5280) were distributed in 16 cages of 4m³ at densities of 45, 70, 95 and 120 fish m-3. The fish were fed with commercial feed with 40% crude protein during all experimental period. Data were subjected to analysis of variance and linear regression to estimate the best fit using 5% probability. The production parameters such as survival, initial weight, final weight, feed conversion and specific growth rate were not influenced by stocking density, as well as the parameters of water quality. However, final biomass, biomass gain and feed intake showed a positive linear relationship with stocking density. Evaluating these results, it was concluded that the stocking density of matrinxã (Brycon orthotaenia) up to 120 fish m-3 represent the best productive performance.

The objective of this study was to evaluate the morphometric growth and weight gain of strains of tilapia (Thai, Red, UFLA and Commercial) by nonlinear models. Initially, 500 male fingerlings of each strain, at 85 (Red and UFLA) and 86 (Thai and Commercial) days of age, were stocked separately in raceways with 56 m³. Twenty fish of each strain were randomly sampled, weighed and measured monthly. Five nonlinear models (Brody, von Bertalanffy, Gompertz, logistic and exponential) were tested, choosing one that best fit to the data. The variables studied were: weight, standard length (SL), head length (HL), height 1 (H1), height 2 (H2), height 3 (H3), first distance (D1), second distance (D2), first width (W1), second width (W2) and third width (W3). The exponential model had the best fit to weight and morphometric data, with the exception of W2, in which the best fitted model was von Bertalanffy. The convergence of the exponential model to data indicates that the cultivation period studied was not enough for the strains to reach maturity weight. The UFLA strain presented the lowest value for parameter "a" (initial weight estimate), 8.71 g, and the highest for parameter k (specific growth rate), 0.0127, when compared with other evaluated strains. However, the highest k of UFLA was not enough to overcome the final weight observed for the Commercial strain (603.1 g), which was higher than all other strains. Regarding the morphometric measurements, the UFLA strain also had the highest k for the variables SL, HL, HH, H1, H2, H3 and D2, and similar k to Commercial and Thai strains for the variables D1 and W3 respectively. The strains differ as to weight gain and morphometric growth.