abstract
- Lightweight model development has emerged as an important study subject in computer vision in response to the need for resource-efficient solutions. These models attempt to strike a balance between model size, computing requirements, and accuracy. They give benefits such as efficient resource use, faster inference times, and improved accessibility. For 3D facial reconstruction models, lightweight architectures present an opportunity for implementation in less demanding hardware, since these algorithms usually rely on powerful processors such as NVIDIA graphic cards. The following research paper provides a benchmark comparison between diverse state-of-the-art lightweight models in a facial reconstruction model, with the aim to reduce its computational complexity so that it can be tested on a mobile device. © 2024, The Author(s), under exclusive license to Springer Nature Switzerland AG.