Abstract:Over the past ten years, tiny Unmanned Aerial Vehicles (UAVs) have exploded in popularity for a variety of aerial monitoring applications, including livestock counting, wildlife tracking in their natural environments, and agricultural region monitoring. When used in conjunction with deep learning, they make it possible to analyze and recognize images automatically. Recently, species population detection and monitoring in remotely sensed data has been made possible with the use of deep learning, an efficient machine learning technology. Animal recognition in satellite and aerial photos is one area where deep learning approaches are finding practical use right now, and this paper intends to give an experimental review of those areas. To simplify the process of animal identification from satellite images, this study presented a new deep learning method called the Learning based Animal Sensing and Classification Model (LASCM). To test how well the method worked, it was cross-validated with the traditional Random Forest (RF) model. The primary obstacles to implementing these deep learning techniques are unbalanced datasets, tiny samples, tiny objects, picture annotation techniques, picture backgrounds, animal counting, evaluation of model performance, and uncertainty calculation. Barely and self-supervised techniques for learning, optimizing either favorable or adverse instances, improving network architecture, and sample annotation methods were all considered as potential answers. The following areas are projected to get increased attention in the next years: video-based detection; detection based on extremely high-resolution satellite images; identification of several species; novel methods for annotation; and the creation of specialized network frameworks and big foundational modelling. The proposed methodology is designed to sort out all these problems and provide an efficient animal identification scheme based on deep learning model from the satellite images.

Bootstrapping Rare Object Detection in High-Resolution Satellite Imagery

State-of-the-Art Deep Learning Methods for Objects Detection in Remote Sensing Satellite Images

Resource-Constrained Simultaneous Detection and Labeling of Objects in High-Resolution Satellite Images

IS-Count: Large-Scale Object Counting from Satellite Images with Covariate-Based Importance Sampling

Exploring Robust Features for Few-Shot Object Detection in Satellite Imagery

Self-Supervised Pretraining and Controlled Augmentation Improve Rare Wildlife Recognition in UAV Images

Improving performance of aircraft detection in satellite imagery while limiting the labelling effort: Hybrid active learning

Satellite Imagery Multiscale Rapid Detection with Windowed Networks

Small or Far Away? Exploiting Deep Super-Resolution and Altitude Data for Aerial Animal Surveillance

Parallel Cascade R-CNN for Object Detection in Remote Sensing Imagery

On-Board Multi-Class Geospatial Object Detection Based on Convolutional Neural Network for High Resolution Remote Sensing Images

Few-Shot Object Detection in Remote Sensing: Lifting the Curse of Incompletely Annotated Novel Objects

Active learning for object detection in high-resolution satellite images

Enhancing Poaching Predictions for Under-Resourced Wildlife Conservation Parks Using Remote Sensing Imagery

Automated aerial animal detection when spatial resolution conditions are varied

Satellite Image Based Animal Identification System Using Deep Learning Assisted Remote Sensing Strategy

Half a Percent of Labels is Enough: Efficient Animal Detection in UAV Imagery Using Deep CNNs and Active Learning

Detecting Mammals in UAV Images: Best Practices to address a substantially Imbalanced Dataset with Deep Learning

Towards Multi-class Object Detection in Unconstrained Remote Sensing Imagery

A hierarchical object detection method in large-scale optical remote sensing satellite imagery using saliency detection and CNN