Ali Swanson | Searcharxiv

arXiv2018

Floating Forests: Quantitative Validation of Citizen Science Data Generated From Consensus Classifications

Large-scale research endeavors can be hindered by logistical constraints limiting the amount of available data. For example, global ecological questions require a global dataset, and traditional sampling protocols are often too inefficient for a small research team to collect an adequate amount of data. Citizen science offers an alternative by crowdsourcing data collection. Despite growing popularity, the community has been slow to embrace it largely due to concerns about quality of data collected by citizen scientists. Using the citizen science project Floating Forests (http://floatingforests.org), we show that consensus classifications made by citizen scientists produce data that is of comparable quality to expert generated classifications. Floating Forests is a web-based project in which citizen scientists view satellite photographs of coastlines and trace the borders of kelp patches. Since launch in 2014, over 7,000 citizen scientists have classified over 750,000 images of kelp forests largely in California and Tasmania. Images are classified by 15 users. We generated consensus classifications by overlaying all citizen classifications and assessed accuracy by comparing to expert classifications. Matthews correlation coefficient (MCC) was calculated for each threshold (1-15), and the threshold with the highest MCC was considered optimal. We showed that optimal user threshold was 4.2 with an MCC of 0.400 (0.023 SE) for Landsats 5 and 7, and a MCC of 0.639 (0.246 SE) for Landsat 8. These results suggest that citizen science data derived from consensus classifications are of comparable accuracy to expert classifications. Citizen science projects should implement methods such as consensus classification in conjunction with a quantitative comparison to expert generated classifications to avoid concerns about data quality.

physics.soc-ph↗

arXiv2017

Automatically identifying, counting, and describing wild animals in camera-trap images with deep learning

Having accurate, detailed, and up-to-date information about the location and behavior of animals in the wild would revolutionize our ability to study and conserve ecosystems. We investigate the ability to automatically, accurately, and inexpensively collect such data, which could transform many fields of biology, ecology, and zoology into "big data" sciences. Motion sensor "camera traps" enable collecting wildlife pictures inexpensively, unobtrusively, and frequently. However, extracting information from these pictures remains an expensive, time-consuming, manual task. We demonstrate that such information can be automatically extracted by deep learning, a cutting-edge type of artificial intelligence. We train deep convolutional neural networks to identify, count, and describe the behaviors of 48 species in the 3.2-million-image Snapshot Serengeti dataset. Our deep neural networks automatically identify animals with over 93.8% accuracy, and we expect that number to improve rapidly in years to come. More importantly, if our system classifies only images it is confident about, our system can automate animal identification for 99.3% of the data while still performing at the same 96.6% accuracy as that of crowdsourced teams of human volunteers, saving more than 8.4 years (at 40 hours per week) of human labeling effort (i.e. over 17,000 hours) on this 3.2-million-image dataset. Those efficiency gains immediately highlight the importance of using deep neural networks to automate data extraction from camera-trap images. Our results suggest that this technology could enable the inexpensive, unobtrusive, high-volume, and even real-time collection of a wealth of information about vast numbers of animals in the wild.

cs.CV↗