RecursiveDet: End-to-End Region-based Recursive Object Detection

TL;DR

RecursiveDet introduces a parameter-sharing recursive decoder with positional encoding for region-based object detection, achieving performance improvements with fewer parameters and minimal additional computation.

Contribution

It proposes a novel recursive decoder with shared parameters and positional encoding, reducing model complexity while enhancing detection accuracy.

Findings

Significant performance boost on mainstream detectors

Fewer model parameters with comparable or better accuracy

Effective use of positional encoding for adaptive proposals

Abstract

End-to-end region-based object detectors like Sparse R-CNN usually have multiple cascade bounding box decoding stages, which refine the current predictions according to their previous results. Model parameters within each stage are independent, evolving a huge cost. In this paper, we find the general setting of decoding stages is actually redundant. By simply sharing parameters and making a recursive decoder, the detector already obtains a significant improvement. The recursive decoder can be further enhanced by positional encoding (PE) of the proposal box, which makes it aware of the exact locations and sizes of input bounding boxes, thus becoming adaptive to proposals from different stages during the recursion. Moreover, we also design centerness-based PE to distinguish the RoI feature element and dynamic convolution kernels at different positions within the bounding box. To validate the effectiveness of the proposed method, we conduct intensive ablations and build the full model on three recent mainstream region-based detectors. The RecusiveDet is able to achieve obvious performance boosts with even fewer model parameters and slightly increased computation cost. Codes are available at https://github.com/bravezzzzzz/RecursiveDet.