Fish Feeding Intensity Assessment in Aquaculture: A New Audio Dataset AFFIA3K and a Deep Learning Algorithm

Abstract

Fish feeding intensity assessment (FFIA) aims to evaluate the change of fish appetite during the feeding process, which is potentially useful in industrial aquaculture. Previous methods are mainly based on computer vision techniques. However, these methods are limited by water refraction and uneven illumination. In this paper, we introduce a new approach for FFIA using audio. We create a new audio dataset for FFIA, namely AFFIA3K, which contains 3000 labelled audio clips of different fish feeding intensity (None, Weak, Medium, Strong). We present a deep learning framework for FFIA, where the audio signal is first transformed into acoustic features, i.e. mel spectrogram, then a convolutional neural network (CNN)-based model is used to classify the fish feeding intensity. Experimental results show that our approach achieves an mAP of 0.74 on the test set of AFFIA3K, and considerably outperforms baseline systems. This indicates the potential of our proposed approach in aquaculture applications.

Augmenting Aquaculture Efficiency through Involutional Neural Networks and Self-Attention for Oplegnathus Punctatus Feeding Intensity Classification from Log Mel Spectrograms

“…The growth, health, and sustainability of aquaculture depend on fish-feeding behavior [ 16 , 19 ]. Researchers can improve fish growth by studying their feeding behavior, while reducing waste and environmental impact.…”

“…They achieved 85.9% accuracy, representing substantial progress toward using deep learning to decode aquaculture-related actions. Cui et al [ 19 ] leveraged Mel Spectrogram features and a convolutional neural network to classify fish feeding behaviors, obtaining a mean average precision of 74%. This measure considers both precision and recall rather than mere accuracy.…”

A real-time feeding decision method based on density estimation of farmed fish

“…One is the fish density analysis dataset (DlouFishDensity), which contains 20 side-shot videos with complete fish feeding behavior, as shown in Figure 11. Another is a data video clip taken by Cui (Cui et al, 2022), which contains videos all taken above the water surface, as shown in Figure 12.…”

“…The videos that meet our requirements have a resolution of 1920*1080 and a duration between 30s-60s. Second, we manually labeled the locations of fish heads on 900 image frames in DlouFishDensity and 300 image frames in the video of Cui (Cui et al, 2022) dataset, and finally labeled the locations of more than 30000 fish heads. The labeling tool uses MATLAB to establish a data_marker labeling applet to label the fish head of each picture in the data set.…”

“…We divided the dataset into a training set and a test set at a ratio of 9:1, with 18 videos for training and 2 videos for testing. From the 18 videos, we extracted 1000 frames of fish school images for manual 4.1.2 The Cui (Cui et al, 2022) dataset…”