Cascaded Dual Vision Transformer for Accurate Facial Landmark Detection

TL;DR

This paper presents a novel dual vision transformer model with long skip connections for facial landmark detection, achieving superior accuracy on multiple benchmarks by modeling geometric relations and preserving low-level features.

Contribution

The paper introduces a dual vision transformer architecture with channel-split ViT and long skip connections, enhancing landmark detection accuracy over previous methods.

Findings

Outperforms state-of-the-art on WFLW, COFW, and 300W benchmarks.

Effectively models geometric relations among landmarks.

Preserves low-level image features for improved prediction.

Abstract

Facial landmark detection is a fundamental problem in computer vision for many downstream applications. This paper introduces a new facial landmark detector based on vision transformers, which consists of two unique designs: Dual Vision Transformer (D-ViT) and Long Skip Connections (LSC). Based on the observation that the channel dimension of feature maps essentially represents the linear bases of the heatmap space, we propose learning the interconnections between these linear bases to model the inherent geometric relations among landmarks via Channel-split ViT. We integrate such channel-split ViT into the standard vision transformer (i.e., spatial-split ViT), forming our Dual Vision Transformer to constitute the prediction blocks. We also suggest using long skip connections to deliver low-level image features to all prediction blocks, thereby preventing useful information from being discarded by intermediate supervision. Extensive experiments are conducted to evaluate the performance of our proposal on the widely used benchmarks, i.e., WFLW, COFW, and 300W, demonstrating that our model outperforms the previous SOTAs across all three benchmarks.