A Multimodal Computer Vision Approach for Wildlife Bird Surveillance using Drones

Remote wildlife monitoring of avian species in difficult-to-access environments has consistently been hypothesized to be a challenge. Previously, wildlife surveillance required conservationists to physically approach nesting grounds, which posed a risk of causing significant stress to the birds, potentially leading them to break their eggs and suffer long-term detrimental effects on their population. During breeding season surveillance is strictly prohibited, further complicating data collection efforts. This study scrutinizes the implementation of an advanced model we developed titled Optimized Rigorous Advanced Cutting-edge model for Leveraging protection to Ecosystems «ORACLE», which encompasses state-of-the-art computer vision techniques for comprehensive remote monitoring of wildlife avian species under inaccessible environments. The goal of this model is to gather collective or individual data on nesting grounds through alternative safer methods. Additionally, we introduce a new innovative methodology for data collection, utilizing multimodal systems to identify specific patterns using Visual Question Answering (VQA) models.