I, with my team has achieved an impressive top 70 ranking and earned a Bronze Award 🥉 by applying state-of-the-art machine learning techniques to a challenging task—bird sound classification. The competition aimed to address a critical ecological challenge: understanding bird populations and their habitats through sound analysis. With over 10,000 bird species worldwide, their vocalizations serve as vital indicators of environmental health and quality of life. However, analyzing vast sound datasets manually is labor-intensive and often incomplete.
For this competition, I harnessed the power of EfficientNet, a cutting-edge neural network architecture, to identify bird sounds within complex soundscape recordings. These recordings presented a unique challenge, as they included not only target bird species but also various background noises, such as anthropogenic sounds and other wildlife calls. My innovative approach and model design allowed me to accurately classify bird vocalizations, contributing to the development of AI-based solutions for ecological conservation.
By securing a top 70 position and a Bronze Award in this competition, I demonstrated my proficiency in machine learning, particularly in the domain of natural language classification, and my commitment to contributing to real-world conservation efforts. This project underscores my dedication to leveraging data science for positive ecological impact and showcases my ability to tackle complex, real-world challenges through innovative AI solutions.
Notebook Link: https://www.kaggle.com/code/dipta007/inference-cbirdcall-2