BridgeTA: Bridging the Representation Gap in Knowledge Distillation via Teacher Assistant for Bird's Eye View Map Segmentation

TL;DR

BridgeTA introduces a lightweight Teacher Assistant network to effectively bridge the representation gap in knowledge distillation for Bird's Eye View map segmentation, improving accuracy without increasing inference costs.

Contribution

It proposes a novel cost-effective distillation framework using a Teacher Assistant network and a theoretical loss grounded in Young's Inequality to enhance knowledge transfer.

Findings

Achieves 4.2% mIoU improvement over baseline

Up to 45% better than other KD methods

Effective on the nuScenes dataset

Abstract

Bird's-Eye-View (BEV) map segmentation is one of the most important and challenging tasks in autonomous driving. Camera-only approaches have drawn attention as cost-effective alternatives to LiDAR, but they still fall behind LiDAR-Camera (LC) fusion-based methods. Knowledge Distillation (KD) has been explored to narrow this gap, but existing methods mainly enlarge the student model by mimicking the teacher's architecture, leading to higher inference cost. To address this issue, we introduce BridgeTA, a cost-effective distillation framework to bridge the representation gap between LC fusion and Camera-only models through a Teacher Assistant (TA) network while keeping the student's architecture and inference cost unchanged. A lightweight TA network combines the BEV representations of the teacher and student, creating a shared latent space that serves as an intermediate representation. To ground the framework theoretically, we derive a distillation loss using Young's Inequality, which decomposes the direct teacher-student distillation path into teacher-TA and TA-student dual paths, stabilizing optimization and strengthening knowledge transfer. Extensive experiments on the challenging nuScenes dataset demonstrate the effectiveness of our method, achieving an improvement of 4.2% mIoU over the Camera-only baseline, up to 45% higher than the improvement of other state-of-the-art KD methods.