FishNet: A Hybrid Deep Learning and Machine Learning Framework for Precise Fish Species Identification

Published: March 21, 2024

This article uses CNNs and Random Forest models to automate fish species identification. The neural network design in Table 3 combines CNN hierarchical feature extraction interpretable ensemble learning, combining their capabilities. study carefully addresses data gathering preparation problems, emphasizing the need for a broad, well-prepared dataset. Model optimization Section C hyperparameter tweaking, regularization, machine learning create balanced effective model. D shows model's resilience varied environmental conditions during recognition execution. 2 displays precision, recall, Fl scores, demonstrating versatility across species. findings advance ecological computer vision offer viable tool regulating fisheries, monitoring, conservation. From matrix of confusion, class-specific metrics, future research, report suggests that automated identification systems can have real-world impact be continuously improved.

Language: Английский

---

Enhanced deep learning models for automatic fish species identification in underwater imagery

Heliyon, Journal Year: 2024, Volume and Issue: 10(15), P. e35217 - e35217

Published: July 27, 2024

Underwater cameras are crucial in marine ecology, but their data management needs automatic species identification. This study proposes a two-stage deep learning approach. First, the Unsharp Mask Filter (UMF) preprocesses images. Then, an enhanced region-based fully convolutional network (R-FCN) detects fish using two-order integrals for position-sensitive score maps and precise region of interest (PS-Pr-RoI) pooling accuracy. The second stage integrates ShuffleNetV2 with Squeeze Excitation (SE) module, forming Improved model, enhancing classification focus. Hyperparameters optimized Enhanced Northern Goshawk Optimization Algorithm (ENGO). improved R-FCN model achieves 99.94 % accuracy, 99.58 precision recall, 99.27 F-measure on Fish4knowledge dataset. Similarly, ENGO-based evaluated same dataset, shows 99.93 99.19 precision, 98.29 98.71 F-measure, highlighting its superior

Language: Английский

---

Probabilistic Model-Based Active Learning with Attention Mechanism for Fish Species Recognition

Published: Sept. 25, 2023

Accurate fish species identification is essential for stock assessments, production management, document ecosystem changes, and protection of endangered species. Image processing computer vision techniques have been widely employed detection, classification, tracking, reducing human efforts in these tasks. However, methods often rely on extensive training data with correct annotations. Annotating many images captured from marine environments poses a significant challenge. This work proposes deep-learning model designed detection classification. The incorporates an attention mechanism named Convolutional Block Attention Module (CBAM) to improve performance. A popular Deep Active Learning approach cost-efficient annotation employed, which selects the most informative samples unlabeled set. proposed method utilizes probabilistic modeling based mixture density networks estimate probability distributions localization classification heads. study uses Southeast Area Monitoring Assessment Program Dataset 2021 (SEAMAPD21). Our compared conventional supervised algorithm. Experimental results demonstrate superior accuracy, achieving mean average precision (mAP) 41.6% minimal labeled data, traditional approaches (mAP-36.7%) that larger datasets. active learning module effectively reduces costs while maintaining excellent accuracy. Overall, our deep proves be highly effective recognition, providing advancements accuracy cost efficiency

Language: Английский

---

A Zero Shot Detection Based Approach for Fish Species Recognition in Underwater Environments

Published: Sept. 25, 2023

Identification of fish species is vital for fisheries management, stock assessments, protection endangered species, and ecosystem management. Image based surveys often deploy video cameras that are used to collect large image datasets reviewed by a human observer identify generate numerical count at each station. One main challenge in labeling or annotating such dataset it requires huge amount time, cost, effort. Recently, general adversarial network (GAN) generative techniques have drawn much attention zero-shot object detection (ZSD) because superior performance localizing simultaneously recognizing objects without training model on unseen (few target) classes. In this work, Fish Species Recognition (ZSD-FR) underwater environments utilized detection. This approach can localize recognize when the not trained "unseen" Generative models like GAN be data with "seen" classes generating class samples depending upon semantics (attributes) learned from seen The results obtained SEAMAPD21 illustrate zero shot successfully transfer knowledge better accuracy.

Language: Английский

---

Active detection for fish species recognition in underwater environments

Published: June 6, 2024

Fish species must be identified for stock assessments, ecosystem monitoring, production management, and the conservation of endangered species. Implementing algorithms fish detection in underwater settings like Gulf Mexico poses a formidable challenge. Active learning, method that efficiently identifies informative samples annotation while staying within budget, has demonstrated its effectiveness context object recent times. In this study, we present an active model designed recognition environments. This can employed as system to effectively lower expense associated with manual annotation. It uses epistemic uncertainty Evidential Deep Learning (EDL) proposes novel module denoted Model Evidence Head (MEH) employs Hierarchical Uncertainty Aggregation (HUA) obtain informativeness image. We conducted experiments using fine-grained extensive dataset reef collected from Mexico, specifically Southeast Area Monitoring Assessment Program Dataset 2021 (SEAMAPD21). The experimental results demonstrate framework achieves better performance on SEAMAPD21 demonstrating favorable balance between data efficiency recognition.

Language: Английский

---