FishNet: Deep Neural Networks for Low-Cost Fish Stock Estimation
Authors:
Moseli Mots'oehli,
Anton Nikolaev,
Wawan B. IGede,
John Lynham,
Peter J. Mous,
Peter Sadowski
Abstract:
Fish stock assessment often involves manual fish counting by taxonomy specialists, which is both time-consuming and costly. We propose FishNet, an automated computer vision system for both taxonomic classification and fish size estimation from images captured with a low-cost digital camera. The system first performs object detection and segmentation using a Mask R-CNN to identify individual fish f…
▽ More
Fish stock assessment often involves manual fish counting by taxonomy specialists, which is both time-consuming and costly. We propose FishNet, an automated computer vision system for both taxonomic classification and fish size estimation from images captured with a low-cost digital camera. The system first performs object detection and segmentation using a Mask R-CNN to identify individual fish from images containing multiple fish, possibly consisting of different species. Then each fish species is classified and the length is predicted using separate machine learning models. To develop the model, we use a dataset of 300,000 hand-labeled images containing 1.2M fish of 163 different species and ranging in length from 10cm to 250cm, with additional annotations and quality control methods used to curate high-quality training data. On held-out test data sets, our system achieves a 92% intersection over union on the fish segmentation task, a 89% top-1 classification accuracy on single fish species classification, and a 2.3cm mean absolute error on the fish length estimation task.
△ Less
Submitted 27 June, 2024; v1 submitted 16 March, 2024;
originally announced March 2024.
Tiebout sorting in online communities
Authors:
John Lynham,
Philip R. Neary
Abstract:
This paper proposes a stylized, dynamic model to address the issue of sorting online. There are two large homogeneous groups of individuals. Everyone must choose between two online platforms, one of which has superior amenities (akin to having superior local public goods). Each individual enjoys interacting online with those from their own group but dislikes being on the same platform as those in…
▽ More
This paper proposes a stylized, dynamic model to address the issue of sorting online. There are two large homogeneous groups of individuals. Everyone must choose between two online platforms, one of which has superior amenities (akin to having superior local public goods). Each individual enjoys interacting online with those from their own group but dislikes being on the same platform as those in the other group. Unlike a Tiebout model of residential sorting, both platforms have unlimited capacity so there are no constraints on cross-platform migration. It is clear how each group would like to sort themselves but, in the presence of the other type, only the two segregated outcomes are guaranteed to be equilibria. Integration on a platform can be supported in equilibrium as long as the platform is sufficiently desirable. If online integration of the two communities is a desired social outcome then the optimal policy is clear: make the preferred platform even more desirable. Revitalizing the inferior platform will never lead to integration and even increases the likelihood the segregation. Finally, integration is more elastic in response to an increase in platform amenities than to reductions in intolerance.
△ Less
Submitted 18 May, 2024; v1 submitted 8 October, 2021;
originally announced October 2021.
