# A deep neural network for multi-species fish detection using multiple acoustic cameras

arXiv: https://arxiv.org/abs/2109.10664 · HTML: https://mlsanity.com/paper/2109.10664/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 20 of its 61 references that are in the map.
- First posted on arXiv 22 Sep 2021.

## Builds on (most cited first, 20 of 20)

- [Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks](https://mlsanity.com/paper/1506.01497/) (2015)
- [PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation](https://mlsanity.com/paper/1612.00593/) (2016)
- [Fully Convolutional Networks for Semantic Segmentation](https://mlsanity.com/paper/1411.4038/) (2014)
- [You Only Look Once: Unified, Real-Time Object Detection](https://mlsanity.com/paper/1506.02640/) (2015)
- [Rich feature hierarchies for accurate object detection and semantic segmentation](https://mlsanity.com/paper/1311.2524/) (2013)
- [Reconciling modern machine learning practice and the bias-variance trade-off](https://mlsanity.com/paper/1812.11118/) (2018)
- [Revisiting Unreasonable Effectiveness of Data in Deep Learning Era](https://mlsanity.com/paper/1707.02968/) (2017)
- [Recognition in Terra Incognita](https://mlsanity.com/paper/1807.04975/) (2018)
- [LSTM Fully Convolutional Networks for Time Series Classification](https://mlsanity.com/paper/1709.05206/) (2017)
- [Optimization for deep learning: theory and algorithms](https://mlsanity.com/paper/1912.08957/) (2019)
- [Deep Learning for Classification of Hyperspectral Data: A Comparative Review](https://mlsanity.com/paper/1904.10674/) (2019)
- [How much data is needed to train a medical image deep learning system to achieve necessary high accuracy?](https://mlsanity.com/paper/1511.06348/) (2015)
- [On the Origin of Deep Learning](https://mlsanity.com/paper/1702.07800/) (2017)
- [Deep Learning Object Detection Methods for Ecological Camera Trap Data](https://mlsanity.com/paper/1803.10842/) (2018)
- [Spatially Supervised Recurrent Convolutional Neural Networks for Visual Object Tracking](https://mlsanity.com/paper/1607.05781/) (2016)
- [An Experimental-based Review of Image Enhancement and Image Restoration Methods for Underwater Imaging](https://mlsanity.com/paper/1907.03246/) (2019)
- [Deep learning with self-supervision and uncertainty regularization to count fish in underwater images](https://mlsanity.com/paper/2104.14964/) (2021)
- [Deep Learning from Shallow Dives: Sonar Image Generation and Training for Underwater Object Detection](https://mlsanity.com/paper/1810.07990/) (2018)
- [Underwater Optical Image Processing: A Comprehensive Review](https://mlsanity.com/paper/1702.03600/) (2017)
- [Deep Learning based Segmentation of Fish in Noisy Forward Looking MBES Images](https://mlsanity.com/paper/2006.09034/) (2020)

## Abstract (arXiv)

Underwater acoustic cameras are high potential devices for many applications in ecology, notably for fisheries management and monitoring. However how to extract such data into high value information without a time-consuming entire dataset reading by an operator is still a challenge. Moreover the analysis of acoustic imaging, due to its low signal-to-noise ratio, is a perfect training ground for experimenting with new approaches, especially concerning Deep Learning techniques. We present hereby a novel approach that takes advantage of both CNN (Convolutional Neural Network) and classical CV (Computer Vision) techniques, able to detect a generic class ''fish'' in acoustic video streams. The pipeline pre-treats the acoustic images to extract 2 features, in order to localise the signals and improve the detection performances. To ensure the performances from an ecological point of view, we propose also a two-step validation, one to validate the results of the trainings and one to test the method on a real-world scenario. The YOLOv3-based model was trained with data of fish from multiple species recorded by the two common acoustic cameras, DIDSON and ARIS, including species of high ecological interest, as Atlantic salmon or European eels. The model we developed provides satisfying results detecting almost 80% of fish and minimizing the false positive rate, however the model is much less efficient for eel detections on ARIS videos. The first CNN pipeline for fish monitoring exploiting video data from two models of acoustic cameras satisfies most of the required features. Many challenges are still present, such as the automation of fish species identification through a multiclass model. 1 However the results point a new solution for dealing with complex data, such as sonar data, which can also be reapplied in other cases where the signal-to-noise ratio is a challenge.