FCOS

Last updated on Feb 23, 2021

FCOS (R-101, 1x, caffe, GN=Y, MS train=N, Tricks=N, DCN=N)

Memory (M) 5500.0
inference time (s/im) 0.0578
File Size 196.16 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time

Architecture Group Normalization, Non Maximum Suppression, ResNet, FPN
MS train N
lr sched 1x
Memory (M) 5500.0
Backbone Layers 101
inference time (s/im) 0.0578
SHOW MORE
SHOW LESS
FCOS (R-101, 2x, caffe, GN=Y, MS train=Y)

Memory (M) 5500.0
inference time (s/im) 0.0578
File Size 196.16 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time

Architecture Group Normalization, Non Maximum Suppression, ResNet, FPN
MS train Y
lr sched 2x
Memory (M) 5500.0
Backbone Layers 101
inference time (s/im) 0.0578
SHOW MORE
SHOW LESS
FCOS (R-50, 1x, caffe, GN=Y, MS train=N, Tricks=N, DCN=N)

Memory (M) 3600.0
inference time (s/im) 0.04405
File Size 123.47 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time

Architecture Group Normalization, Non Maximum Suppression, ResNet, FPN
MS train N
lr sched 1x
Memory (M) 3600.0
Backbone Layers 50
inference time (s/im) 0.04405
SHOW MORE
SHOW LESS
FCOS (R-50, 1x, caffe, GN=Y, MS train=N, Tricks=Y, DCN=N)

Memory (M) 3700.0
Backbone Layers 50
File Size 123.47 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time

Architecture Group Normalization, Non Maximum Suppression, ResNet, FPN
MS train N
lr sched 1x
Memory (M) 3700.0
Backbone Layers 50
SHOW MORE
SHOW LESS
FCOS (R-50, 1x, caffe, GN=Y, MS train=N, Tricks=Y, DCN=Y)

Memory (M) 3800.0
Backbone Layers 50
File Size 127.28 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time

Architecture Non Maximum Suppression, Group Normalization, Deformable Convolution, FPN, ResNet
MS train N
lr sched 1x
Memory (M) 3800.0
Backbone Layers 50
SHOW MORE
SHOW LESS
FCOS (R-50, 2x, caffe, GN=Y, MS train=Y)

Memory (M) 2600.0
inference time (s/im) 0.04367
File Size 123.47 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time

Architecture Group Normalization, Non Maximum Suppression, ResNet, FPN
MS train Y
lr sched 2x
Memory (M) 2600.0
Backbone Layers 50
inference time (s/im) 0.04367
SHOW MORE
SHOW LESS
FCOS (X-101, 2x, pytorch, GN=Y, MS train=Y)

Memory (M) 10000.0
inference time (s/im) 0.10309
File Size 344.95 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time

Architecture Non Maximum Suppression, Group Normalization, ResNeXt, FPN
MS train Y
lr sched 2x
Memory (M) 10000.0
Backbone Layers 101
inference time (s/im) 0.10309
SHOW MORE
SHOW LESS
README.md

FCOS: Fully Convolutional One-Stage Object Detection

Introduction

[ALGORITHM]

@article{tian2019fcos,
  title={FCOS: Fully Convolutional One-Stage Object Detection},
  author={Tian, Zhi and Shen, Chunhua and Chen, Hao and He, Tong},
  journal={arXiv preprint arXiv:1904.01355},
  year={2019}
}

Results and Models

Backbone Style GN MS train Tricks DCN Lr schd Mem (GB) Inf time (fps) box AP Config Download
R-50 caffe Y N N N 1x 3.6 22.7 36.6 config model | log
R-50 caffe Y N Y N 1x 3.7 - 38.7 config model | log
R-50 caffe Y N Y Y 1x 3.8 - 42.3 config model | log
R-101 caffe Y N N N 1x 5.5 17.3 39.1 config model | log
Backbone Style GN MS train Lr schd Mem (GB) Inf time (fps) box AP Config Download
R-50 caffe Y Y 2x 2.6 22.9 38.5 config model | log
R-101 caffe Y Y 2x 5.5 17.3 40.8 config model | log
X-101 pytorch Y Y 2x 10.0 9.7 42.6 config model | log

Notes:

  • The X-101 backbone is X-101-64x4d.
  • Tricks means setting norm_on_bbox, centerness_on_reg, center_sampling as True.
  • DCN means using DCNv2 in both backbone and head.

Results

Object Detection on COCO minival

Object Detection on COCO minival
MODEL BOX AP
FCOS (X-101, 2x, pytorch, GN=Y, MS train=Y) 42.6
FCOS (R-50, 1x, caffe, GN=Y, MS train=N, Tricks=Y, DCN=Y) 42.3
FCOS (R-101, 2x, caffe, GN=Y, MS train=Y) 40.8
FCOS (R-101, 1x, caffe, GN=Y, MS train=N, Tricks=N, DCN=N) 39.1
FCOS (R-50, 1x, caffe, GN=Y, MS train=N, Tricks=Y, DCN=N) 38.7
FCOS (R-50, 2x, caffe, GN=Y, MS train=Y) 38.5
FCOS (R-50, 1x, caffe, GN=Y, MS train=N, Tricks=N, DCN=N) 36.6