Impact of the State-of-the-Art Methods on Camera Trap Image Classification

Result code in IS VaVaI
<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F49777513%3A23520%2F24%3A43973114" target="_blank" >RIV/49777513:23520/24:43973114 - isvavai.cz</a>
Result on the web
<a href="https://svk.fav.zcu.cz/download/proceedings_svk_2024.pdf" target="_blank" >https://svk.fav.zcu.cz/download/proceedings_svk_2024.pdf</a>
DOI - Digital Object Identifier
—

Result language
angličtina
Original language name
Impact of the State-of-the-Art Methods on Camera Trap Image Classification
Original language description
Camera traps are valuable assets in ecological research. They are commonly used to estimate wildlife populations, species distribution, and interactions. In many cases, the data are still processed manually, which is extremely time-consuming, given the relatively high number of operated camera traps and their continuous data flow. Therefore, a concerted effort is being made to automate this process using machine learning and computer vision.This article compares Camera Trap Image Classification approaches with an adaptation of the Multi-Modal methods- BLIP by Li, et. al. (2022) and ChatGPT sourced from Ruu3f (2023). Even though the Multi-Modal methods have never seen the data used, they generate almost 1/3 correct predictions. However, the standard approaches based on the BEiTv2 classifier are noticeably more accurate, achieving up to 68.2% of accuracy on the CCT20 dataset.
Czech name
—
Czech description
—

Publication year
2024
Confidentiality
S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Similar results(10)