VFNet

open-mmlab / mmdetection

Last updated on Feb 23, 2021
VFNet (R-101, 1x, pytorch, DCN=N)
lr sched 1x
Backbone Layers 101
File Size 198.54 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time
Architecture VFNet, ResNet, FCOS, Varifocal Loss
lr sched 1x
Backbone Layers 101
SHOW MORE
SHOW LESS
VFNet (R-101, 2x, pytorch, DCN=N)
lr sched 2x
Backbone Layers 101
File Size 198.54 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time
Architecture VFNet, ResNet, FCOS, Varifocal Loss
lr sched 2x
Backbone Layers 101
SHOW MORE
SHOW LESS
VFNet (R-101, 2x, pytorch, DCN=Y)
lr sched 2x
Backbone Layers 101
File Size 206.39 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time
Architecture Varifocal Loss, FCOS, Deformable Convolution, FPN, VFNet, ResNet
lr sched 2x
Backbone Layers 101
SHOW MORE
SHOW LESS
VFNet (R-50, 1x, pytorch, DCN=N)
lr sched 1x
Backbone Layers 50
File Size 125.79 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time
Architecture VFNet, ResNet, FCOS, Varifocal Loss
lr sched 1x
Backbone Layers 50
SHOW MORE
SHOW LESS
VFNet (R-50, 2x, pytorch, DCN=N)
lr sched 2x
Backbone Layers 50
File Size 125.79 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time
Architecture VFNet, ResNet, FCOS, Varifocal Loss
lr sched 2x
Backbone Layers 50
SHOW MORE
SHOW LESS
VFNet (R-50, 2x, pytorch, DCN=Y)
lr sched 2x
Backbone Layers 50
File Size 129.60 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time
Architecture Varifocal Loss, FCOS, Deformable Convolution, FPN, VFNet, ResNet
lr sched 2x
Backbone Layers 50
SHOW MORE
SHOW LESS
VFNet (X-101-32x4d, 2x, pytorch, DCN=Y)
lr sched 2x
Backbone Layers 101
File Size 212.45 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time
Architecture ResNeXt, Varifocal Loss, FCOS, Deformable Convolution, FPN, VFNet
lr sched 2x
Backbone Layers 101
SHOW MORE
SHOW LESS
VFNet (X-101-64x4d, 2x, pytorch, DCN=Y)
lr sched 2x
Backbone Layers 101
File Size 377.25 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time
Architecture ResNeXt, Varifocal Loss, FCOS, Deformable Convolution, FPN, VFNet
lr sched 2x
Backbone Layers 101
SHOW MORE
SHOW LESS
README.md

VarifocalNet: An IoU-aware Dense Object Detector

Introduction

[ALGORITHM]

VarifocalNet (VFNet) learns to predict the IoU-aware classification score which mixes the object presence confidence and localization accuracy together as the detection score for a bounding box. The learning is supervised by the proposed Varifocal Loss (VFL), based on a new star-shaped bounding box feature representation (the features at nine yellow sampling points). Given the new representation, the object localization accuracy is further improved by refining the initially regressed bounding box. The full paper is available at: https://arxiv.org/abs/2008.13367.

Learning to Predict the IoU-aware Classification Score.

Citing VarifocalNet

@article{zhang2020varifocalnet,
  title={VarifocalNet: An IoU-aware Dense Object Detector},
  author={Zhang, Haoyang and Wang, Ying and Dayoub, Feras and S{\"u}nderhauf, Niko},
  journal={arXiv preprint arXiv:2008.13367},
  year={2020}
}

Results and Models

Backbone Style DCN MS train Lr schd Inf time (fps) box AP (val) box AP (test-dev) Config Download
R-50 pytorch N N 1x - 41.6 41.6 config model | log
R-50 pytorch N Y 2x - 44.5 44.8 config model | log
R-50 pytorch Y Y 2x - 47.8 48.0 config model | log
R-101 pytorch N N 1x - 43.0 43.6 config model | log
R-101 pytorch N Y 2x - 46.2 46.7 config model | log
R-101 pytorch Y Y 2x - 49.0 49.2 config model | log
X-101-32x4d pytorch Y Y 2x - 49.7 50.0 config model | log
X-101-64x4d pytorch Y Y 2x - 50.4 50.8 config model | log

Notes:

  • The MS-train scale range is 1333x[480:960] (range mode) and the inference scale keeps 1333x800.
  • DCN means using DCNv2 in both backbone and head.
  • Inference time will be updated soon.
  • More results and pre-trained models can be found in VarifocalNet-Github

Results

Object Detection on COCO minival
MODEL BOX AP
VFNet (X-101-64x4d, 2x, pytorch, DCN=Y) 50.4
VFNet (X-101-32x4d, 2x, pytorch, DCN=Y) 49.7
VFNet (R-101, 2x, pytorch, DCN=Y) 49.0
VFNet (R-50, 2x, pytorch, DCN=Y) 47.8
VFNet (R-101, 2x, pytorch, DCN=N) 46.2
VFNet (R-50, 2x, pytorch, DCN=N) 44.5
VFNet (R-101, 1x, pytorch, DCN=N) 43.0
VFNet (R-50, 1x, pytorch, DCN=N) 41.6
Contact us on: hello@paperswithcode.com . Papers With Code is a free resource with all data licensed under CC-BY-SA.
Terms Data policy Cookies policy