Aggregated Residual Transformations for Deep Neural Networks

Saining XieRoss GirshickPiotr DollárZhuowen TuKaiming He

   Papers with code   Abstract  PDF

We present a simple, highly modularized network architecture for image classification. Our network is constructed by repeating a building block that aggregates a set of transformations with the same topology... (read more)

Benchmarked Models

RANK
MODEL
REPO
CODE RESULT
PAPER RESULT
ε-REPRODUCED
BUILD
1
ResNeXt-101
(64x4d)
80.4%
--
2
ResNeXt-101
(32x4d)
80.0%
--
3
ResNeXt-50
(32x4d)
79.2%
--
4
ResNeXt-101-32x8d
79.1%
79.3%
5
ResNeXt-101-32x8d
79.1%
79.3%
7
ResNeXt-101 32x8d
79.1%
--
8
ResNeXt-50 32x4d
77.5%
--
9
ResNeXt-26
(32x4d)
75.9%
--
10
ResNeXt-26
(32x2d)
73.4%
--
11
ResNeXt-14
(32x4d)
69.7%
--
12
ResNeXt-14
(16x4d)
68.1%
--
13
ResNeXt-14
(32x2d)
67.4%
--