Fish Images Research Articles

Re-identification of fish individuals of undulate skate via deep learning within a few-shot context

Individual re-identification is critical to track population changes in order to assess status, being particularly relevant in species with conservation concerns and difficult access like marine organisms. For this, we propose photo-identification via deep learning as a non-invasive technique to discriminate between individuals of the undulate skate (Raja undulata). Nevertheless, accruing enough training samples might be difficult to achieve in the case of underwater fish images. We develop a novel methodology based on a siamese neural network that incorporates statistical fundamentals as motivation to overcome the few-shot context. Our work provides a hands-on experience and highlights on pitfalls when trying to apply photo-identification in a limited scenario, concerning both data quantity and quality, yet providing remarkable results over the test set including recaptures, where the model is capable of correctly identifying the 70% of the individuals. The findings of this study can be of strong impact for the research teams becoming familiar with deep learning approaches, as it can be easily extended to re-identify individuals of other marine species of interest from a conservation or exploitation point of view.

Large Scale Fish Images Classification and Localization using Transfer Learning and Localization Aware CNN Architecture

Building an automatic fish recognition and detection system for large-scale fish classes is helpful for marine researchers and marine scientists because there are large numbers of fish species. However, it is quite difficult to build such systems owing to the lack of data imbalance problems and large number of classes. To solve these issues, we propose a transfer learning-based technique in which we use Efficient-Net, which is pre-trained on ImageNet dataset and fine-tuned on QuT Fish Database, which is a large scale dataset. Furthermore, prior to the activation layer, we use Global Average Pooling (GAP) instead of dense layer with the aim of averaging the results of predictions along with having more information compared to the dense layer. To check the validity of our model, we validate our model on the validation set which achieves satisfactory results. Also, for the localization task, we propose an architecture that consists of localization aware block, which captures localization information for better prediction and residual connections to handle the over-fitting problem. Actually, the residual connections help the layer to combine missing information with the relevant one. In addition, we use class weights and Focal Loss (FL) to handle class imbalance problems along with reducing false predictions. Actually, class weights assign less weights to classes having fewer instances and large weights to classes having more number of instances. During the localization, the qualitative assessment shows that we achieve 57% Mean Intersection Over Union (IoU) on testing data, and the classification results show 75% precision, 70% recall, 78% accuracy and 74% F1-Score for 468 fish species.

Underwater Inhomogeneous Light Field Based on Improved Convolutional Neural Net Fish Image Recognition

Underwater Inhomogeneous Light Field Based on Improved Convolutional Neural Net Fish Image Recognition

Detection of the Types of Consumable Saltwater Fish in the Coastal Area of Likupang

In the coastal area of Likupang, many types of saltwater fish can be consumed, such as tuna and skipjack. Yet, there are also types of saltwater fish that cannot be consumed or protected by the government, such as Napoleon fish and sea kingfish. Thus, this research aimed to build a desktop application that can automatically classify consumable and non-consumable saltwater fish species more accurately and promptly using a suitable image recognition method like the Convolutional Neural Network (CNN). CNN has abilities to distinguish images by recognizing several pixels in a two-dimensional image and RGB (Red, Green, Blue) colors which are then converted into a matrix with various values, making it easier for the system to recognize the two-dimensional image. By using 40% test data (143 images) and 60% training data (213 images), test accuracy in identifying and classifying images of consumable fish, non-consumable fish, and non-fish images with each percentage of 94%, 98%, and 95% respectively.

Identifikasi Kualitas Kesegaran Ikan Menggunakan Algoritma K-Nearest Neighbor Berdasarkan Ekstraksi Ciri Warna Hue, Saturation, dan Value (HSV)

Fish has a very high nutritional content and is needed by the human body, such as a protein. With the increasing production, and need for consumption of goodand fresh fish, irresponsible sellers take advantage of this situation by selling fish that are not fit for consumption, such as fish that are not fresh (rotten), fish that contain chlorine and formalin which can be detrimental to consumers. The purpose of this study was to determine how accurate the identification of fish freshness quality using the extraction of Hue, Saturation, and Value (HSV) color characteristics. The research method used is K-Nearest Neighbor (KNN) and is classified into several parts, namely, data collection techniques, needs analysis, design, training, and then testing. The image sample data used in this study amounted to 240 images consisting of fresh and non-fresh fish images, which will then be divided into training data and test data. The training data sample amounted to 220 images with a division of 110 fresh fish images and 110 non-fresh fish images, while the test data sample totaled 20 images with a division of 10 fresh fish images and 10 non-fresh fish images. Analysis of color features is carried out on the gills and head or the area around the eyes of the fish using Red, Green, and Blue (RGB) colors, which will be converted into Hue, Saturation, and Value (HSV) color spaces for the extraction and training processes to obtain results. The results showed that the use of HSV color character extraction was successfully applied with an accuracy value in the training of 94.09% and testing of 90%

MATSYASTRA - An Automated Fish Species Identification using Teachable Machine Services

Generally, only feature values obtained from photos are used to identify fish species. But, it is challenging to identify fish species based on an image alone because fish of the same species can have varying hues or seem quite similar to other species. Additionally, it can be a tedious task that might lead to wrong predictions. Since various fish species exist, it is difficult to determine a fish without a proper model. Fast-growing computing and sensing technologies have improved most embedded systems, which help us solve more complicated algorithms. The main challenge is to perceive and analyze corresponding information for better judgment. An advanced system with better computing power can facilitate identifying fish species. Using the Teachable machine, a web-based tool for creating machine learning models, we can ensure that this application gives accurate results in classifying various fish species. An application that uses machine learning to identify fish categories is developed in this study by capturing images of fish and identifying their categories. In addition to providing fish information, this app also connects users with other fishermen, gives feedback on the fish, display catch logs, supports multilingual display of data, fish focused advisory chatbot, and market value information. User dashboards allow users to sign up, create profiles, scan, and identify their catches. This mobile application ensures the data integrity and confidentiality of the user’s data. The overall performance of the application is responsive and user friendly.

Identification of Freshwater Fish Types Using Linear Discriminant Analysis (LDA) Algorithm

Fish as aquatic animals have several physiological mechanisms that land animals do not have. Differences in habitat cause fish to adapt to environmental conditions, for example as animals that live in water, both in fresh and marine waters. The number of species or types of freshwater fish means knowledge of the types of freshwater fish. Identification of freshwater fish images is useful for the community, because the types of freshwater fish have different nutritional content, prices and processing for each type. Likewise for cultivators, identification of freshwater fish species can be useful for providing fish handling and management because each fish has a different cultivation method. The purpose of this study was to identify freshwater fish species using the Linear Discriminant Analysis (LDA) algorithm based on color feature extraction using HSV. The LDA algorithm has the ability to reduce dimensions by dividing data into several groups by maximizing the distance between groups that are different or more. To make the identification process easier, color feature extraction with HSV can be used to extract a variety of information from the color in the image. Based on the results of the accuracy test, it produces a value of 84.5%, which is included in the good category.

Rapid non-destructive detection of foreign bodies in fish based on terahertz imaging and spectroscopy

ObjectiveFood safety is a matter of social stability and healthy development of the national economy. The processing technology of fish complicated and tedious and it is inclined to have fish bone residues or mixed with various foreign bodies during the processing. It is difficult to detect these foreign bodies by traditional manual and visual detection methods, which poses a serious threat to the health of people. In this case, this paper aims to explore the feasibility of rapid and non-destructive detection of endogenous foreign bodies (fish bones) and exogenous foreign bodies (metal, plastic, wooden toothpicks) in fish based on terahertz imaging and spectroscopy. MethodsFirstly, the transmission terahertz spectral images of fish samples are acquired, and the terahertz spectral information of the region of interest is extracted. Secondly, in order to improve the accuracy of the model and reduce the computational effort, three algorithms including Competitive Adaptive Reweighting Sampling (CARS), Uninformative Variable Elimination (UVE) and Successive Projections Algorithm (SPA) are used to extract the spectral features. Thirdly, three qualitative discriminant models including Partial Least Squares Discriminant Analysis (PLS-DA), Linear Discriminant Analysis (LDA) and Support Vector Machine (SVM) are established to detect and identify foreign bodies in fish. Finally, in order to realize the visual expression of Foreign Bodies, the terahertz images of fish are segmented by binarization threshold. ResultsThe model is optimized to obtain the best SVM model after CARS feature extraction with the accuracy rate of 99.56%. Meanwhile, the terahertz images of the samples at 0.1–4.0 THz are visualized and analyzed. It is found that the foreign bodies in fish are clearly outlined, and the shape and size of foreign bodies in fish can be well distinguished from terahertz images. ConclusionThis study has proved that the terahertz imaging and spectroscopy technology can achieve rapid and non-destructive detection of endogenous foreign bodies (fish bones) and exogenous foreign bodies (metal, plastic, wooden toothpicks) in fish. It provides relevant basis and theoretical reference for rapid non-destructive detection of foreign bodies in food, which is of great significance for food process monitoring and quality control.

Freshwater Fish Classification Based on Image Representation Using K-Nearest Neighbor Method

Indonesia is a maritime and agricultural country with enormous world fishery potential. The large variety of fish is often confusing for ordinary people in recognizing types of fish, especially freshwater fish. It was stated that the types of freshwater fish often consumed by the Indonesian people are bawal (pomfret), betutu, gabus (cork), gurame (carp), mas (goldfish), lele (catfish), mujaer (tilapia), patin (asian catfish), tawes, and nila (tilapia nilotica). Some fish types have similar shapes, so it is tricky to tell them apart. Meanwhile, in the digitalization era today, Artificial Intelligence (AI)-based technology has become a demand in all areas of life. It is overgrowing, not apart from the fisheries sector. Therefore, in this study, the K-Nearest Neighbor (KNN) method was applied as one of the methods in AI to identify and classify freshwater fish species based on their images. The KNN method classifies new data into specific classes based on the distance between the new data and the closest k data through the learning process. This KNN model is built by preparing the dataset stages, separating the dataset into data-train and data-test with a ratio of 70%:30%, then building and testing the model. The dataset is freshwater fish images, totaling 100 images from 10 freshwater fish types. Model testing is done by measuring performance using a confusion matrix. Based on the test results, the model has an accuracy performance of 70%. Thus, KNN can be used as a model to identify freshwater fish species based on their image.

A Scalable Open-Source Framework for Machine Learning-Based Image Collection, Annotation and Classification: A Case Study for Automatic Fish Species Identification

Citizen science platforms, social media and smart phone applications enable the collection of large amounts of georeferenced images. This provides a huge opportunity in biodiversity and ecological research, but also creates challenges for efficient data handling and processing. Recreational and small-scale fisheries is one of the fields that could be revolutionised by efficient, widely accessible and machine learning-based processing of georeferenced images. Most non-commercial inland and coastal fisheries are considered data poor and are rarely assessed, yet they provide multiple societal benefits and can have substantial ecological impacts. Given that large quantities of georeferenced fish images are being collected by fishers every day, artificial intelligence (AI) and computer vision applications offer a great opportunity to automate their analyses by providing species identification, and potentially also fish size estimation. This would deliver data needed for fisheries management and fisher engagement. To date, however, many AI image analysis applications in fisheries are focused on the commercial sector, limited to specific species or settings, and are not publicly available. In addition, using AI and computer vision tools often requires a strong background in programming. In this study, we aim to facilitate broader use of computer vision tools in fisheries and ecological research by compiling an open-source user friendly and modular framework for large-scale image storage, handling, annotation and automatic classification, using cost- and labour-efficient methodologies. The tool is based on TensorFlow Lite Model Maker library, and includes data augmentation and transfer learning techniques applied to different convolutional neural network models. We demonstrate the potential application of this framework using a small example dataset of fish images taken through a recreational fishing smartphone application. The framework presented here can be used to develop region-specific species identification models, which could potentially be combined into a larger hierarchical model.

Представления о рыбах в традиционной культуре лесных юкагиров (по материалам языка, этнографии и фольклора)

The study aims to describe the traditional ideas of the Forest Yukaghir about fish based on the materials of language, ethnography and folklore. Scientific novelty lies in the systematisation of these data, introduction of new information on the ethnography and folklore of the Yukaghir into scientific use. As a result of the study, it has been shown that linguistic and ethnographic materials on the Forest Yukaghir indicate the great importance of fishing in their life. In the non-fairytale and fairytale prose of the Yukagir related to the ichthyofauna, mythological representations associated with the theme of birth are found, the reasons for the acquisition of certain features by fish are explained, folklore and mythological characteristics of individual fish are given; in small genres of folklore, there is an indication of the images of fish in fairy tales, a comparison based on an external feature; in song folklore reflecting the worldview and way of life of fishermen, fish are a marker of ancestral territories, a symbol of natural abundance.

A Robust Feature Construction for Fish Classification Using Grey Wolf Optimizer

Abstract The low quality of the collected fish image data directly from its habitat affects its feature qualities. Previous studies tended to be more concerned with finding the best method rather than the feature quality. This article proposes a new fish classification workflow using a combination of Contrast-Adaptive Color Correction (NCACC) image enhancement and optimization-based feature construction called Grey Wolf Optimizer (GWO). This approach improves the image feature extraction results to obtain new and more meaningful features. This article compares the GWO-based and other optimization method-based fish classification on the newly generated features. The comparison results show that GWO-based classification had 0.22% lower accuracy than GA-based but 1.13 % higher than PSO. Based on ANOVA tests, the accuracy of GA and GWO were statistically indifferent, and GWO and PSO were statistically different. On the other hand, GWO-based performed 0.61 times faster than GA-based classification and 1.36 minutes faster than the other.

Robust detection of farmed fish by fusing YOLOv5 with DCM and ATM

Accurate object detection of farmed fish is vital for fish counting and measuring. However, the blurring, occlusion, and varying density of underwater fish images from aquaculture ponds result in low precision and poor robustness. This paper proposes a robust detection model called DCM-ATM-YOLOv5, in which the deformable convolution module (DCM) and adaptive threshold module (ATM) are integrated into YOLOv5. The farmed fish detection model is based on YOLOv5, considering its high precision and real-time capabilities. To achieve higher precision on blurred targets, the DCM is applied to offset the sampling locations and to focus more on fish so that the features of blurry fish are enhanced and the background is suppressed. The ATM is introduced to alleviate missed detections resulting from fixed thresholds by generating appropriate thresholds for different densities and predicting the thresholds dynamically, improving the robustness of the proposed model. Two experiments based on data from aquaculture ponds were designed to test the performance of the proposed model. The average precision and recall of DCM-ATM-YOLOv5 were 97.53% and 98.09%, respectively. Compared with the average precision and recall of YOLOv5, those of DCM-ATM-YOLOv5 were improved by 1.92% and 1.97%, respectively. The robustness of the proposed model was evaluated based on data with different densities and settings. The detection results obtained from datasets of varying densities and settings demonstrated that the detection performance of DCM-ATM-YOLOv5 was still superior to that of YOLOv5, indicating that the proposed model has significant stability and robustness. The experimental results showed that DCM-ATM-YOLOv5 has high precision and strong robustness in genuine aquaculture settings and can provide a feasible solution for the detection of farmed fish.

A novel automatic detection method for abnormal behavior of single fish using image fusion

Fish behavior detection is extremely important for farmers to get information on life indicators of fish, which could be useful to prevent disease outbreaks, predict water quality changes, and improve fish welfare. However, conventional fish disease detection does not meet real-time detection requirements and could have an irreversible influence on fish. Additionally, abnormal detection based on school behavior does not allow for the detection of early abnormal behavior in single fish. To solve this problem, a novel method of abnormal behavior detection based on image fusion was proposed. Firstly, the outline information of the moving object was extracted based on image processing technology. Secondly, the position information of the fish image was enhanced using mosaic image fusion. Finally, bidirectional feature pyramid network, coordinate attention block, and spatial pyramid pooling were added to YOLOv5, which was named BCS-YOLOv5. And compared with the other two typical models, the BCS-YOLOv5 based image fusion achieved the best accuracy with an average accuracy of 96.69% at 45 frames per second in four typical behavior datasets. The proposed method not only improves the extraction of location information but also quantitatively detects similar anomalous behavior, which meets the demand for real-time detection of fish abnormal behavior in aquaculture.

Impact evaluation of deep learning on image segmentation for automatic bluefin tuna sizing

This paper evaluates the impact of using deep learning techniques in an automatic fish sizing process. Automatic fish sizing with a non-invasive approach involves working with different views of the fish's body and changing environments, being the stage of extraction of individuals in the image and the quality of the segmentation essential to obtain good sizing measurements. The goal of this work is to improve the results and functionality achieved in our previous studies with conventional segmentation methods based on local thresholding, where different limitations were observed, mainly the necessity of parameters tuning and a high computational cost. The number of detections must also increase significantly to increase the reliability of the statistical results. An approach using convolutional neural networks is proposed for fish detection and segmentation in videos acquired under real conditions, which eliminates the engineering procedure of parameter adjustment and generalises the solution for fish segmentation to deal with different environmental conditions (illumination and water turbidity) and background variability. The results show that the fish sizing procedure is enhanced thanks to the improvement in fish image instance segmentation. In particular, the number of fish measurements increases by up to 2.45 times when using Mask R-CNN and the PointRend module, thus increasing the accuracy of the fish length estimation, and the number of measurements per minute of computing time increases by up to 3.5 times. Our proposal obtains highly accurate fish length estimations in juvenile bluefin tuna based on a stereoscopic vision system and a deformable model of the fish’s silhouette, both from the ventral and dorsal perspectives. An important improvement is achieved by applying CNN, as demonstrated by the number of segmented instances, the time required to segment an instance, and the accuracy of the fish sizing achieved.

Fish image recognition method based on multi-layer feature fusion convolutional network

Solving the problem of fish image classification is important to conserve fish diversity. This conundrum can be addressed by developing a new fish image classification method based on deep learning by training data with complex backgrounds. To this end, this paper proposes a fusion model, referred to as Tripmix-Net. The backbone network of the proposed model primarily consists of multiscale parallel and improved residual networks that are connected in an alternate manner, and network fusion is used to integrate the information that is extracted from shallow and deep layers. Experiments conducted on the 15-category WildFish fish image dataset verified the efficacy of the proposed Tripmix-Net for classifying same-genus fish images with complex backgrounds. The model achieved an accuracy of 95.31%, which is a significant improvement over traditional methods. The proposed approach serves as a new concept for the fine-grained image classification of fish against complex backgrounds.

A Super-resolution Reconstruction Method for Underwater Fish Images with Feedback Networks

<p>Video surveillance is commonly used for production process monitoring in aquaculture. Because underwater video images are prone to blurring and unclear detail texture, this paper proposes a super-resolution network reconstruction method with a feedback network. The reconstruction effect is enhanced by a feedback network in the process of mapping low-resolution images to high-resolution images, which is implemented with a constrained recurrent neural network (RNN) to process deep feedback information. However, the shallow features are fed to the feedback net-work module to generate deeper features through multiple upsampling and downsampling and progressive refinement, which give the network structure an early reconstruction capability and are conducive to more realistically reconstructing high-resolution images. A course learning strategy is introduced to make the network applicable to more complex tasks and improve its robust-ness. We construct a training set of 800 images and validation and test sets of 10 images each using images of marine net fish as the research object. The method of this paper is validated on the public dataset Set14 and a self-built dataset. The proposed method outperforms other methods in both subjective and objective evaluations on the public and self-built datasets. Good foundation for high-definition monitoring of aquaculture is laid. </p> <p> </p>

Образы змеи и рыбы в архитектурной декорации христианских храмов Кавказа в эпоху средневековья: некоторые размышления

The article deals with the images of fish and snake in the architectural decoration of the Christian churches in Caucasus in 10 – 14 centuries. The theme of the dragon fighters is not in the focus of the research. The research is dedicated to the images of snake (without paws and fangs, in contrast to dragon-snake) when represented in heraldic position, in the scenes of the prophet Jonas devoured by the sea-monster and in isolated position. It is also dedicated to the image of the fish. The earliest representations of the snakes and dragon-snakes appear on the facades of the Christian churches in Caucasus as early as the Xth century on the territory of the Northern Caucasus and Georgia. Later they are turned in hybrid creatures, dragon-snake (with paws and fangs, sometimes covered with fish scales). Two heads of snake on the earliest representations as well as cirсles on its body underline the apotropaic meaning of the image. The appearance of the snake (not the dragon-snake) on the territory in question could be explained by the influence of the local archaeological culture (Cobano-Colchis). Being represented on the facades of the Christian churches they function as apotropaia. The appearance of the isolated images of fishes in the architectural decoration could also be influenced by the local archaeological culture. The way of the sketching the scales on its body (on the relief from Anacopia) proves its biological identification as trout (the widest spread fish in the mountains of Caucasus). The image of the fish carved on the facades of the Christian churches could also have apotropaic function.

Combination of Feature Extractions for Classification of Coral Reef Fish Types Using Backpropagation Neural Network

Feature extraction is important to obtain information in digital images, where feature extraction results are used in the classification process. The success of a study to classify digital images is highly dependent on the selection of the feature extraction method used, from several studies providing a combination of feature extraction solutions to produce a more accurate classification. Classifying the types of marine fish is done by identifying fish based on special characteristics, and it can be through a description of the shape, fish body pattern, color, or other characteristics. This study aimed to classify coral reef fish species based on the characteristics contained in fish images using Backpropagation Neural Network (BPNN) method. Data used in this research was collected directly from Bunaken National Marine Park (BNMP) in Indonesia. The first stage was to extract shape features using the Geometric Invariant Moment (GIM) method, texture features using Gray Level Co-occurrence Matrix (GLCM) method, and color feature extraction using Hue Saturation Value (HSV) method. The third value of feature extraction was used as input for the next stage, namely the classification process using the BPNN method. The test results using 5-fold cross-validation found that the lowest test accuracy was 85%, the highest was 100%, and the average was 96%. This means that the intelligent model derived from the combination of the three feature extraction methods implemented in the BPNN training algorithm is very good for classifying coral reef fish.

Fish Freshness Identification Using Machine Learning: Performance Comparison of k-NN and Naïve Bayes Classifier

Fish is one of the food sources that should be examined for freshness before being consumed. The consumption of rotten fish can cause various diseases. The rotten fish have changed color on the gills, skin, flesh, and eyes and have a pungent odour. Fish freshness can be assessed using a variety of conventional methods, but these methods have limitations, such as requiring relatively expensive equipment and trained personnel and being destructive. We use machine learning because it is non-destructive, reduces costs, and is easy to use. This study aims to identify the freshness of fish using k-nearest neighbor (k-NN) and Naïve Bayes (NB) classification methods based on the fish-eye image. The features used in the classification process are RGB and GLCM. The research stages consist of the fish collection process, image acquisition and class division, preprocessing and ROI detection, feature extraction and dataset split, and the classification process. The results show that the k-NN method has better performance than NB with average accuracy, precision, recall, specificity, and AUC of 0.97, 0.97, 0.97, 0.97, and 0.97, respectively.