Implicit Feature Pyramid Network for Object Detection
This work provides an incremental improvement in object detection performance for computer vision researchers and practitioners by enhancing existing FPN architectures.
This paper introduces an implicit feature pyramid network (i-FPN) that models FPN transformations using an implicit function. It significantly improves object detection performance on the MS COCO dataset, boosting mAP by +3.4 on RetinaNet, +3.2 on Faster-RCNN, +3.5 on FCOS, +4.2 on ATSS, and +3.2 on AutoAssign compared to ResNet-50-FPN baselines.
In this paper, we present an implicit feature pyramid network (i-FPN) for object detection. Existing FPNs stack several cross-scale blocks to obtain large receptive field. We propose to use an implicit function, recently introduced in deep equilibrium model (DEQ), to model the transformation of FPN. We develop a residual-like iteration to updates the hidden states efficiently. Experimental results on MS COCO dataset show that i-FPN can significantly boost detection performance compared to baseline detectors with ResNet-50-FPN: +3.4, +3.2, +3.5, +4.2, +3.2 mAP on RetinaNet, Faster-RCNN, FCOS, ATSS and AutoAssign, respectively.