Abstract:Recognising animals based on distinctive body patterns, such as stripes, spots, or other markings, in night images is a complex task in computer vision. Existing methods for detecting animals in images often rely on colour information, which is not always available in night images, posing a challenge for pattern recognition in such conditions. Nevertheless, recognition at night-time is essential for most wildlife, biodiversity, and conservation applications. The SPOTS-10 dataset was created to address this challenge and to provide a resource for evaluating machine learning algorithms in situ. This dataset is an extensive collection of grayscale images showcasing diverse patterns found in ten animal species. Specifically, SPOTS-10 contains 50,000 32 x 32 grayscale images, divided into ten categories, with 5,000 images per category. The training set comprises 40,000 images, while the test set contains 10,000 images. The SPOTS-10 dataset is freely available on the project GitHub page: <a class="link-external link-https" href="https://github.com/Amotica/SPOTS-10.git" rel="external noopener nofollow">this https URL</a> by cloning the repository.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: the complex task of identifying animals based on their unique body surface patterns (such as stripes, spots or other markings) in nocturnal images. Existing methods usually rely on color information, but in nocturnal images, color information is often unavailable, which poses a challenge to pattern recognition. Therefore, nocturnal animal identification is crucial for applications such as wildlife conservation, biodiversity research and ecological monitoring. ### Specific background of the problem 1. **Lack of color information in nocturnal images**: Most existing animal identification methods rely on color information, but in images taken at night, due to insufficient light, color information is not obvious or completely absent, resulting in poor performance of these methods under nocturnal conditions. 2. **Complexity of the natural environment**: Camera traps in the wild are often used to capture the natural behaviors of animals, but the environments where these devices are located are complex and changeable, including dense vegetation and constantly changing lighting conditions, which cause the animal parts in the images to be occluded and increase the difficulty of identification. 3. **Requirements of specific application scenarios**: Identifying animals at night is very important for applications such as wildlife conservation, biodiversity research and ecological monitoring. For example, understanding the behavioral patterns, population dynamics and ecosystem health of nocturnal animals can provide key data for conservation work. ### Solution To address the above challenges, the author created a dataset named **SPOTS - 10**. This dataset contains 50,000 32×32 - pixel grayscale images, covering 10 different species of animals, with 5,000 images for each category. The main features of this dataset are as follows: - **Grayscale images**: All images are grayscale images, simulating the effect of nocturnal camera - trap shooting, ensuring that the model can be trained and tested in the absence of color information. - **Diverse animal patterns**: The images in the dataset show the unique patterns of various animals, such as stripes, spots, etc., which help the model to learn and recognize these features. - **Benchmark test**: The SPOTS - 10 dataset is not only used to develop new machine - learning algorithms, but also can be used as a benchmark dataset to evaluate the performance of existing algorithms in nocturnal images. By creating the SPOTS - 10 dataset, the author hopes to promote the development in the fields of computer vision and machine learning, especially in nocturnal animal identification, thereby providing more powerful tools for wildlife conservation and ecological research.

SPOTS-10: Animal Pattern Benchmark Dataset for Machine Learning Algorithms

A dataset of ground-dwelling nocturnal fauna for object detection and classification

OpenAnimalTracks: A Dataset for Animal Track Recognition

AnimalTrack: A Benchmark for Multi-Animal Tracking in the Wild

Optimizing the automated recognition of individual animals to support population monitoring

ANIMAL-SPOT enables animal-independent signal detection and classification using deep learning

AnimalWeb: A Large-Scale Hierarchical Dataset of Annotated Animal Faces

Batch-Mask: An automated Mask R-CNN workflow to isolate non-standard biological specimens for color pattern analysis

PetFace: A Large-Scale Dataset and Benchmark for Animal Identification

BuckTales : A multi-UAV dataset for multi-object tracking and re-identification of wild antelopes

Animal Detection and Classification from Camera Trap Images Using Different Mainstream Object Detection Architectures

AP-10K: A Benchmark for Animal Pose Estimation in the Wild

dopanim: A Dataset of Doppelganger Animals with Noisy Annotations from Multiple Humans

WildlifeReID-10k: Wildlife re-identification dataset with 10k individual animals

CatFLW: Cat Facial Landmarks in the Wild Dataset

Animal Recognition and Identification with Deep Convolutional Neural Networks for Automated Wildlife Monitoring

PatternNet: A benchmark dataset for performance evaluation of remote sensing image retrieval

MammalNet: A Large-scale Video Benchmark for Mammal Recognition and Behavior Understanding

Insect Identification in the Wild: The AMI Dataset

Understanding the Impact of Training Set Size on Animal Re-identification

Pose Recognition in the Wild: Animal pose estimation using Agglomerative Clustering and Contrastive Learning