LADet: A Light-weight and Adaptive Network for Multi-scale Object Detection

Jiaming Zhou, Yuqiao Tian, Weicheng Li, Rui Wang, Zhongzhi Luan, Depei Qian
Proceedings of The Eleventh Asian Conference on Machine Learning, PMLR 101:912-923, 2019.

Abstract

Scale variation is one of the most significant challenges for object detection task. In comparison with previous one-stage object detectors that simply make feature pyramid network deeper without consideration of speed, we propose a novel one-stage object detector called LADet, which consists of two parts, Adaptive Feature Pyramid Module(AFPM) and Light-weight Classification Function Module(LCFM). Adaptive Feature Pyramid Module generates complementary semantic information for each level feature map by jointly utilizing multi-level feature maps from backbone network, which is different from the top-down manner. Light-weight Classification Function Module is able to exploit more type of anchor boxes without a dramatic increase of parameters because of the utilization of interleaved group convolution. Extensive experiments on PASCAL VOC and MS COCO benchmark demonstrate that our model achieves a better trade-off between accuracy and efficiency over the comparable state-of-the-art detection methods.

Cite this Paper

BibTeX
@InProceedings{pmlr-v101-zhou19a, title = {LADet: A Light-weight and Adaptive Network for Multi-scale Object Detection}, author = {Zhou, Jiaming and Tian, Yuqiao and Li, Weicheng and Wang, Rui and Luan, Zhongzhi and Qian, Depei}, booktitle = {Proceedings of The Eleventh Asian Conference on Machine Learning}, pages = {912--923}, year = {2019}, editor = {Lee, Wee Sun and Suzuki, Taiji}, volume = {101}, series = {Proceedings of Machine Learning Research}, month = {17--19 Nov}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v101/zhou19a/zhou19a.pdf}, url = {https://proceedings.mlr.press/v101/zhou19a.html}, abstract = {Scale variation is one of the most significant challenges for object detection task. In comparison with previous one-stage object detectors that simply make feature pyramid network deeper without consideration of speed, we propose a novel one-stage object detector called LADet, which consists of two parts, Adaptive Feature Pyramid Module(AFPM) and Light-weight Classification Function Module(LCFM). Adaptive Feature Pyramid Module generates complementary semantic information for each level feature map by jointly utilizing multi-level feature maps from backbone network, which is different from the top-down manner. Light-weight Classification Function Module is able to exploit more type of anchor boxes without a dramatic increase of parameters because of the utilization of interleaved group convolution. Extensive experiments on PASCAL VOC and MS COCO benchmark demonstrate that our model achieves a better trade-off between accuracy and efficiency over the comparable state-of-the-art detection methods.} }
Endnote
%0 Conference Paper %T LADet: A Light-weight and Adaptive Network for Multi-scale Object Detection %A Jiaming Zhou %A Yuqiao Tian %A Weicheng Li %A Rui Wang %A Zhongzhi Luan %A Depei Qian %B Proceedings of The Eleventh Asian Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2019 %E Wee Sun Lee %E Taiji Suzuki %F pmlr-v101-zhou19a %I PMLR %P 912--923 %U https://proceedings.mlr.press/v101/zhou19a.html %V 101 %X Scale variation is one of the most significant challenges for object detection task. In comparison with previous one-stage object detectors that simply make feature pyramid network deeper without consideration of speed, we propose a novel one-stage object detector called LADet, which consists of two parts, Adaptive Feature Pyramid Module(AFPM) and Light-weight Classification Function Module(LCFM). Adaptive Feature Pyramid Module generates complementary semantic information for each level feature map by jointly utilizing multi-level feature maps from backbone network, which is different from the top-down manner. Light-weight Classification Function Module is able to exploit more type of anchor boxes without a dramatic increase of parameters because of the utilization of interleaved group convolution. Extensive experiments on PASCAL VOC and MS COCO benchmark demonstrate that our model achieves a better trade-off between accuracy and efficiency over the comparable state-of-the-art detection methods.
APA
Zhou, J., Tian, Y., Li, W., Wang, R., Luan, Z. & Qian, D.. (2019). LADet: A Light-weight and Adaptive Network for Multi-scale Object Detection. Proceedings of The Eleventh Asian Conference on Machine Learning, in Proceedings of Machine Learning Research 101:912-923 Available from https://proceedings.mlr.press/v101/zhou19a.html.

Related Material