Underwater Object Detection Using TC-YOLO with Attention Mechanisms (original) (raw)
Related papers
CNN-based YOLOv3 Comparison for Underwater Object Detection
International Journal of Electrical & Electronic Systems Research (IEESR), 2021
Object detection that deals with identifying and locating object is one of area that integrate from the advancement in machine learning and computer vision. Modern object detection which carried out supervised learning utilizes Convo-lutional Neural Network (CNN) as the backbone of the detection architecture which is significant for underwater object detection as the underwater images are usually low in quality and blurry. Single stage detection such as You Only Look Once (YOLO) is one the famous object detection model that is prominent among researchers due to high performance in accuracy and processing speed. However, YOLO has many versions where the current incremental improvement model of YOLOv3 has been widely used by researchers to solve different types of problem related to object detection. Therefore, there is a need to explore the trade-off relationship between the processing speed and precision of each YOLO model. In the study, two different open source underwater datasets were used in four different YOLOv3 models namely as YOLOv3-SPP, YOLOv3-Tiny, YOLOv3-Tiny-PRN and the original YOLOv3 in order to study their performance based on metrics evaluation of precision and processing speed (FPS). The result shows that YOLOv3-SPP proved to be the best in terms of precision while YOLOv3-Tiny-PRN lead in terms of execution speed. So, this study shows that YOLOv3 model is highly significant to be implemented and able to accurately detect underwater objects with haze and low-light environment. This study can help researchers and industry in determining the best YOLOv3 model specifically for detection of the underwater images and its application.
Optimizing the Hyperparameter Tuning of YOLOv5 for Underwater Detection
IEEE Access
This study optimized the latest YOLOv5 framework, including its subset models, with training on different datasets that differed in image contrast and cloudiness to assess model performances based on quantitative metrics and image processing speed. The hyperparameter in the feature-extraction phase was configured based on the learning rate and momentum and further improved based on the adaptive moment estimation (ADAM) optimizer and the function reducing-learning-rate-on-plateau to optimize the model's training scheme. The optimized YOLOv5s achieved a better performance, with a mean average precision of 98.6% and a high inference speed of 106 frames per second. The ADAM optimizer with a detailed learning rate (0.0001) and momentum (0.99) fine-tuning yielded a sufficient convergence rate (0.69% at 55th epoch) to assist YOLOv5s in attaining a more precise detection for underwater objects. INDEX TERMS Image processing speed, object recognition, optimization model, tuning hyper-parameter, underwater imaging.
International journal for research in applied science and engineering technology ijraset, 2020
: Detecting underwater objects is challenged by various kinds of aspect ratios, object size,material colour, cluttered backgrounds, and in particular,not defined orientations. In our paper, we are using a Probabilistic Neural Network (PNN) features from layers which are combined to perform orientation robust aerial underwater object detection. We explore the essential characteristics of PNN as well as corelate the extracted features to the principle of disengaging feature learning. An image segmentation based approach is used to localize ROIs of different aspect ratios, and furtherly ROIs are classified into positives or negative using a DCNN features. On inquiring the two datasets collected from Google Earth, we illustrate that the proposed aerial underwater object detection approach is simple and easy process. Fast and robust underwater object detection in aerial images is potentially applicable in traffic surveillance, emergency, remote sensing and large scale image content analysis.
Robust Underwater Fish Detection Using an Enhanced Convolutional Neural Network
International Journal of Image, Graphics and Signal Processing, 2021
Underwater Object Detection is one of the most challenging and unexplored domains in this area of Computer Vision. The proposed research refines the image enhancement of underwater imagery by proposing an improvement of already existing tools for underwater Object detection. The comparative study clearly depicts the enhancement of the proposed method with respect to the existing methods for underwater object detection. Moreover, a framework for detection of underwater organisms such as fishes are proposed, which will act as the steppingstone for reserving the ecosystem of the whole fish community. Mostly the object detection using deep learning has been the prime goal to this research and the comparison between other preexisting methods are compared at the end. As a result, techniques that are already well established will be used for overall enhancement of those images. Through this enhancement and through finding a healthy environment for their breeding ground, the extinction of selected species of fishes is can be diminished and decreased. All this is carried out by overcoming difficulties underwater through a novel technique that can be integrated into an Underwater Autonomous Vehicle and can be classified as robust in nature. Robustness will depend on three important factors in this research, first is accuracy, then fast and lastly being upgradeable. The proposed method is a modified VGGNet-16, which is trained using the ImageCLEF FISH_TS dataset. The overall result provides an accuracy of 96.4% which surpasses all its predecessors.
Underwater Object Classification and Detection: first results and open challenges
ArXiv, 2022
This work reviews the problem of object detection in underwater environments. We analyse and quantify the shortcomings of conventional state-of-the-art (SOTA) algorithms in the computer vision community when applied to this challenging environment, as well as providing insights and general guidelines for future research efforts. First, we assessed if pretraining with the conventional ImageNet is beneficial when the object detector needs to be applied to environments that may be characterised by a different feature distribution. We then investigate whether two-stage detectors yields to better performance with respect to single-stage detectors, in terms of accuracy, intersection of union (IoU), floating operation per second (FLOPS), and inference time. Finally, we assessed the generalisation capability of each model to a lower quality dataset to simulate performance on a real scenario, in which harsher conditions ought to be expected. Our experimental results provide evidence that und...
Underwater object detection using Invert Multi-Class Adaboost with deep learning
2020 International Joint Conference on Neural Networks (IJCNN)
In recent years, deep learning based methods have achieved promising performance in standard object detection. However, these methods lack sufficient capabilities to handle underwater object detection due to these challenges: (1) Objects in real applications are usually small and their images are blurry, and (2) images in the underwater datasets and real applications accompany heterogeneous noise. To address these two problems, we first propose a novel neural network architecture, namely Sample-WeIghted hyPEr Network (SWIPENet), for small object detection. SWIPENet consists of high resolution and semanticrich Hyper Feature Maps which can significantly improve small object detection accuracy. In addition, we propose a novel sampleweighted loss function which can model sample weights for SWIPENet, which uses a novel sample re-weighting algorithm, namely Invert Multi-Class Adaboost (IMA), to reduce the influence of noise on the proposed SWIPENet. Experiments on two underwater robot picking contest datasets URPC2017 and URPC2018 show that the proposed SWIPENet+IMA framework achieves better performance in detection accuracy against several state-of-the-art object detection approaches.
An Analysis on Water Objects Detection Techniques
International Journal for Research in Applied Science and Engineering Technology IJRASET, 2020
Water object detection is the process of identifying various objects either on the surface of the water or under water through images or videos. The objects to be detected are like floats, marine species, ships, pipelines etc. In this article, an extensive survey has been made on different strategies developed to detect and recognize underwater objects and objects on the water surface. Various methods have been proposed by numerous scientists to detect water targets based on image processing, neural networks, deep learning methods like faster R-CNN, YOLO, adaptive filtering schemes, background subtraction methods etc. The analysed methods can be used in several areas including aquatic study, maintaining and fixing damages of underwater structures.
This paper presents an efficient method which can be used for underwater object detection system automation. The most important thing in underwater image processing is selection of processing domain (like RGB, Gray-scale, R, G, B etc) and filter. The RGB image is captured by a underwater waterproof camera or taken from database from internet. Underwater image processing for object detection is a system which loads a image, pre-processes the image, filters and scales the image to find the object.
Fast and Accurate Fish Detection Design with Improved YOLO-v3 Model and Transfer Learning
International Journal of Advanced Computer Science and Applications, 2020
Object Detection is one of the problematic Computer Vision (CV) problems with countless applications. We proposed a real-time object detection algorithm based on Improved You Only Look Once version 3 (YOLOv3) for detecting fish. The demand for monitoring the marine ecosystem is increasing day by day for a vigorous automated system, which has been beneficial for all of the researchers in order to collect information about marine life. This proposed work mainly approached the CV technique to detect and classify marine life. In this paper, we proposed improved YOLOv3 by increasing detection scale from 3 to 4, apply k-means clustering to increase the anchor boxes, novel transfer learning technique, and improvement in loss function to improve the model performance. We performed object detection on four fish species custom datasets by applying YOLOv3 architecture. We got 87.56% mean Average Precision (mAP). Moreover, comparing to the experimental analysis of the original YOLOv3 model with the improved one, we observed the mAP increased from 87.17% to 91.30. It showed that improved version outperforms than the original YOLOv3 model.
Deep Learning on Underwater Marine Object Detection: A Survey
Deep learning, also known as deep machine learning or deep structured learning based techniques, have recently achieved tremendous success in digital image processing for object detection and classification. As a result, they are rapidly gaining popularity and attention from the computer vision research community. There has been a massive increase in the collection of digital imagery for the monitoring of underwater ecosystems, including seagrass meadows. This growth in image data has driven the need for automatic detection and classification using deep neural network based classifiers. This paper systematically describes the use of deep learning for underwater imagery analysis within the recent past. The analysis approaches are categorized according to the object of detection , and the features and deep learning architectures used are highlighted. It is concluded that there is a great scope for automation in the analysis of digital sea-bed imagery using deep neural networks, especially for the detection and monitoring of seagrass.