拟投顶会论文的详细提纲 Transformer-based Multi-view Human Mesh Recovery with Occlusion Handling

Title: Transformer-based Multi-view Human Mesh Recovery with Occlusion Handling

Abstract:

This paper addresses the challenging problem of 3D human mesh recovery from multi-view images in the presence of occlusions. Existing methods often struggle with accurately reconstructing human meshes when body parts are obstructed from certain viewpoints. To overcome this limitation, we propose a novel Transformer-based framework that effectively leverages multi-view information and explicitly handles occlusions. Our approach employs a two-stage architecture: first, a Transformer encoder extracts multi-view features while explicitly modeling occlusions through a learned occlusion attention mechanism. Second, a Transformer decoder aggregates the mul