Generalised deep learning model for semi-automated length measurement of fish in stereo-BRUVS

Abstract

Assessing the health of fish populations relies on determining the length of fish in sample species subsets, in conjunction with other key ecosystem markers; thereby, inferring overall health of communities. Despite attempts to use artificial intelligence (AI) to measure fish, most measurement remains a manual process, often necessitating fish being removed from the water. Overcoming this limitation and potentially harmful intervention by measuring fish without disturbance in their natural habitat would greatly enhance and expedite the process. Stereo baited remote underwater video systems (stereo-BRUVS) are widely used as a non-invasive, stressless method for manually counting and measuring fish in aquaculture, fisheries and conservation management. However, the application of deep learning (DL) to stereo-BRUVS image processing is showing encouraging progress towards replacing the manual and labour-intensive task of precisely locating the heads and tails of fish with computer-vision-based algorithms. Here, we present a generalised, semi-automated method for measuring the length of fish using DL with near-human accuracy for numerous species of fish. Additionally, we combine the DL method with a highly precise stereo-BRUVS calibration method, which uses calibration cubes to ensure precision within a few millimetres in calculated lengths. In a human versus DL comparison of accuracy, we show that, although DL commonly slightly over-estimates or under-estimates length, with enough repeated measurements, the two values average and converge to the same length, demonstrated by a Pearson correlation coefficient (r) of 0.99 for n=3954 measurement in ‘out-of-sample’ test data. We demonstrate, through the inclusion of visual examples of stereo-BRUVS scenes, the accuracy of this approach. The head-to-tail measurement method presented here builds on, and advances, previously published object detection for stereo-BRUVS. Furthermore, by replacing the manual process of four careful mouse clicks on the screen to precisely locate the head and tail of a fish in two images, with two fast clicks anywhere on that fish in those two images, a significant reduction in image processing and analysis time is expected. By reducing analysis times, more images can be processed; thereby, increasing the amount of data available for environmental reporting and decision making.

Simultaneous, vision-based fish instance segmentation, species classification and size regression

“… Giordano, Palazzo & Spampinato (2016) focuses on fish behaviour analysis from underwater videos. Marrable et al (2023) proposes a semi-automated method for measuring the length of fish using deep learning with near-human accuracy from stereo underwater video systems. Álvarez-Ellacuría et al (2020) propose the use of a deep convolutional network (Mask R-CNN) for unsupervised length estimation from images of European hake boxes collected at the fish market.…”

Tibia-YOLO: An AssistedDetection System Combined with Industrial CT Equipment for Leg Diseases in Broilers

“…The average error of the blade length was 4.6%, and the average error of the blade width was 5.7%. Marrable, Daniel et al [19] proposed a semi-automatic measurement method based on deep learning for a three-dimensional decoy remote underwater video system. The results proved that the measured Pearson correlation coefficient (R) was 0.99.…”