Snapshot Serengeti, high-frequency annotated camera trap images of 40 mammalian species in an African savanna

By A. Swanson, Margaret Kosmala, Chris J. Lintott, Robert Simpson, Arfon M. Smith, Craig Packer. 2015

524 citations recorded by the source corpus.

Camera traps can be used to address large-scale questions in community ecology by providing systematic data on an array of wide-ranging species. We deployed 225 camera traps across 1,125 km2 in Serengeti National Park, Tanzania, to evaluate spatial and temporal inter-species dynamics. The cameras have operated continuously since 2010 and had accumulated 99,241 camera-trap days and produced 1.2 million sets of pictures by 2013. Members of the general public classified the images via the citizen-science website www.snapshotserengeti.org. Multiple users viewed each image and recorded the species, number of individuals, associated behaviours, and presence of young. Over 28,000 registered users contributed 10.8 million classifications. We applied a simple algorithm to aggregate these individual classifications into a final ‘consensus’ dataset, yielding a final classification for each image and a measure of agreement among individual answers. The consensus classifications and raw imagery provide an unparalleled opportunity to investigate multi-species dynamics in an intact ecosystem and a valuable resource for machine-learning and computer-vision research. Design Type(s) observation design • time series design • species comparison design Measurement Type(s) mammalian species abundance Technology Type(s) Camera-trap Factor Type(s) Sample Characteristic(s) Serengeti National Park • savanna Design Type(s) observation design • time series design • species comparison design Measurement Type(s) mammalian species abundance Technology Type(s) Camera-trap Factor Type(s) Sample Characteristic(s) Serengeti National Park • savanna Machine-accessible metadata file describing the reported data (ISA-Tab format)

Subfields: Ecological Modeling, Ecology, Social Psychology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Ecological Modeling

In Ecology

In Social Psychology

Citations in the corpus

Recognition in Terra Incognita (2018, background): “The Snapshot Serengeti dataset [21] is a large, multi-year camera trap dataset collected at 225 locations in a small region of the African savanna.”
Camera-trap images segmentation using multi-layer robust principal component analysis (2017, unknown): “Another example is the Snapshot Serengeti database [35] where only 26.”
Automatically identifying, counting, and describing wild animals in camera-trap images with deep learning (2017, methodology): “The SS project is the world’s largest camera-trap project published to date, with 225 camera traps running continuously in Serengeti National Park, Tanzania, since 2011 (1).”
A review of camera trapping for conservation behaviour research (2017, unknown): “Emerging large-scale camera monitoring initiatives, such as Snapshot Serengeti (www.snapshotserengeti.org; Swanson et al. 2015) and Wildcam Gorongosa (www.wildcamgorongosa.org), demonstrate the benefits of this approach.”
Towards automatic wild animal monitoring: Identification of animal species in camera-trap images using very deep convolutional neural networks (2016, methodology): “Before the releasing of the Snapshot Serengeti dataset [6] there was no publicly available dataset to work with and for benchmarking.”
Machine learning for image based species identification (2018, background): “Examples highlighting the wealth of images captured by researchers and the public are the iNaturalist (675,000 images of 5,000 species, Van Horn et al., 2017) and the Zooniverse (1.2 million images of 40 species; Swanson et al., 2015) databases.”
Machine learning to classify animal species in camera trap images: applications in ecology (2018, methodology): “A common limitation is these methods lead to a large accumulation of images which must be first classified in order to be used in ecological studies (Niedballa, Sollmann, Courtiol, & Wilting, 2016; Swanson et al., 2015).”
A review of factors to consider when using camera traps to study animal behavior to inform wildlife ecology and conservation (2020, unknown): “It is then possible to assess degrees of interobserver reliability (Martin & Bateson, 1986) and aggregate classifications from multiple participants to improve accuracy in resulting datasets (Hsing et al., 2018; Swanson et al., 2015).”
A Survey of Deep Active Learning (2020, methodology): “Object detection [178] VOC [59], Kitti [77] – [110, 154] SS [215], eMML [67], NACTI1, CCT2, UAV3 Biodiversity survey [63] KITTI [78] Autonomous driving [140] NEU-DET [209] Defect detection Semantic segmentation [74, 232, 233, 243] SPIM [46], Confocal[47], LIDC-IDRI [12], MICCAI, Lymph node [252] Bio”
A computer vision for animal ecology. (2018, background): “from iNaturalist (675,000 images of 5,000 species, Van Horn et al., 2017) and Zooniverse (1.2 million images of 40 species; Swanson et al., 2015), highlight the growth in high-quality images captured by researchers and the public.”
Automatically identifying wild animals in camera trap images with deep learning (2017, methodology): “Unless otherwise speciﬁed, all images like that in (b) are from the Snapshot Serengeti dataset (9).”
The iWildCam 2019 Challenge Dataset (2019, background): “Automating camera trap labeling is not a new challenge for the computer vision community [1, 3–10, 12–17].”
Identifying animal species in camera trap images using deep learning and citizen science (2018, methodology): “Norouzzadeh et al. (2018) reported impressive accuracies using the Snapshot Serengeti dataset which comprises 3.2 million images (Swanson et al., 2015)—much more than most camera trap datasets as observed on Zooniverse.”
Efficient Pipeline for Camera Trap Image Review (2019, methodology): “The computer vision community has been making strides towards automating the species classification challenge in camera traps [1, 2, 4–16], but it has proven difficult to to apply models trained in one region to images collected in different geographic areas.”