FoveaBox: Beyond Anchor-based Object Detector

Tao KongFuchun SunHuaping LiuYuning JiangJianbo Shi

   Papers with code   Abstract  PDF

We present FoveaBox, an accurate, flexible and completely anchor-free framework for object detection. While almost all state-of-the-art object detectors utilize the predefined anchors to enumerate possible locations, scales and aspect ratios for the search of the objects, their performance and generalization ability are also limited to the design of anchors... (read more)

Benchmarked Models

RANK
MODEL
REPO
CODE RESULT
PAPER RESULT
ε-REPRODUCED
BUILD
1
FoveaBox
(ResNet-101, align, LR…)
0.419
--
2
FoveaBox
(ResNet-50, align, LR …)
0.401
--
3
FoveaBox
(ResNet-101, align, LR…)
0.394
--
4
FoveaBox
(ResNet-101-FPN, 800x8…)
0.385
0.389
5
FoveaBox
(ResNet-101, LR 1x)
0.385
--
6
FoveaBox
(ResNet-50, align, LR …)
0.379
--
7
FoveaBox
(ResNet-50, LR 2x)
0.369
0.371
8
FoveaBox
(ResNet-50, LR 1x)
0.365
--