SABL

open-mmlab / mmdetection

Last updated on Feb 23, 2021
SABL Cascade R-CNN (R-101-FPN, 1x, MS train=N)
lr sched 1x
Backbone Layers 101
File Size 346.67 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time
Architecture RPN, Non-Maximum Suppression, FPN, Cascade R-CNN, SABL, ResNet, RoIAlign
MS train N
lr sched 1x
Backbone Layers 101
SHOW MORE
SHOW LESS
SABL Cascade R-CNN (R-50-FPN, 1x, MS train=N)
lr sched 1x
Backbone Layers 50
File Size 273.97 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time
Architecture RPN, Non-Maximum Suppression, FPN, Cascade R-CNN, SABL, ResNet, RoIAlign
MS train N
lr sched 1x
Backbone Layers 50
SHOW MORE
SHOW LESS
SABL Faster R-CNN (R-101-FPN, 1x, MS train=N)
lr sched 1x
Backbone Layers 101
File Size 234.00 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time
Architecture Softmax, RPN, Convolution, FPN, Non-Maximum Suppression, RoIPool, SABL, ResNet
MS train N
lr sched 1x
Backbone Layers 101
SHOW MORE
SHOW LESS
SABL Faster R-CNN (R-50-FPN, 1x, MS train=N)
lr sched 1x
Backbone Layers 50
File Size 161.30 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time
Architecture Softmax, RPN, Convolution, FPN, Non-Maximum Suppression, RoIPool, SABL, ResNet
MS train N
lr sched 1x
Backbone Layers 50
SHOW MORE
SHOW LESS
SABL RetinaNet (R-101-FPN, 1x, GN=N, MS train=N)
lr sched 1x
Backbone Layers 101
File Size 212.35 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time
Architecture FPN, Non-Maximum Suppression, SABL, ResNet, Focal Loss
MS train N
lr sched 1x
Backbone Layers 101
SHOW MORE
SHOW LESS
SABL RetinaNet (R-101-FPN, 1x, GN=Y, MS train=N)
lr sched 1x
Backbone Layers 101
File Size 212.36 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time
Architecture Group Normalization, FPN, Non-Maximum Suppression, SABL, ResNet, Focal Loss
MS train N
lr sched 1x
Backbone Layers 101
SHOW MORE
SHOW LESS
SABL RetinaNet (R-101-FPN, 2x, GN=Y, MS train=Y (480~960))
lr sched 2x
Backbone Layers 101
File Size 212.36 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time
Architecture Group Normalization, FPN, Non-Maximum Suppression, SABL, ResNet, Focal Loss
MS train Y (480~960)
lr sched 2x
Backbone Layers 101
SHOW MORE
SHOW LESS
SABL RetinaNet (R-101-FPN, 2x, GN=Y, MS train=Y (640~800))
lr sched 2x
Backbone Layers 101
File Size 212.36 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time
Architecture Group Normalization, FPN, Non-Maximum Suppression, SABL, ResNet, Focal Loss
MS train Y (640~800)
lr sched 2x
Backbone Layers 101
SHOW MORE
SHOW LESS
SABL RetinaNet (R-50-FPN, 1x, GN=N, MS train=N)
lr sched 1x
Backbone Layers 50
File Size 139.65 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time
Architecture FPN, Non-Maximum Suppression, SABL, ResNet, Focal Loss
MS train N
lr sched 1x
Backbone Layers 50
SHOW MORE
SHOW LESS
SABL RetinaNet (R-50-FPN, 1x, GN=Y, MS train=N)
lr sched 1x
Backbone Layers 50
File Size 139.66 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time
Architecture Group Normalization, FPN, Non-Maximum Suppression, SABL, ResNet, Focal Loss
MS train N
lr sched 1x
Backbone Layers 50
SHOW MORE
SHOW LESS
README.md

Side-Aware Boundary Localization for More Precise Object Detection

Introduction

[ALGORITHM]

We provide config files to reproduce the object detection results in the ECCV 2020 Spotlight paper for Side-Aware Boundary Localization for More Precise Object Detection.

@inproceedings{Wang_2020_ECCV,
    title = {Side-Aware Boundary Localization for More Precise Object Detection},
    author = {Jiaqi Wang and Wenwei Zhang and Yuhang Cao and Kai Chen and Jiangmiao Pang and Tao Gong and Jianping Shi and Chen Change Loy and Dahua Lin},
    booktitle = {ECCV},
    year = {2020}
}

Results and Models

The results on COCO 2017 val is shown in the below table. (results on test-dev are usually slightly higher than val). Single-scale testing (1333x800) is adopted in all results.

Method Backbone Lr schd ms-train box AP Config Download
SABL Faster R-CNN R-50-FPN 1x N 39.9 config model | log
SABL Faster R-CNN R-101-FPN 1x N 41.7 config model | log
SABL Cascade R-CNN R-50-FPN 1x N 41.6 config model | log
SABL Cascade R-CNN R-101-FPN 1x N 43.0 config model | log
Method Backbone GN Lr schd ms-train box AP Config Download
SABL RetinaNet R-50-FPN N 1x N 37.7 config model | log
SABL RetinaNet R-50-FPN Y 1x N 38.8 config model | log
SABL RetinaNet R-101-FPN N 1x N 39.7 config model | log
SABL RetinaNet R-101-FPN Y 1x N 40.5 config model | log
SABL RetinaNet R-101-FPN Y 2x Y (640~800) 42.9 config model | log
SABL RetinaNet R-101-FPN Y 2x Y (480~960) 43.6 config model | log

Results

Object Detection on COCO minival
MODEL BOX AP
SABL RetinaNet (R-101-FPN, 2x, GN=Y, MS train=Y (480~960)) 43.6
SABL Cascade R-CNN (R-101-FPN, 1x, MS train=N) 43.0
SABL RetinaNet (R-101-FPN, 2x, GN=Y, MS train=Y (640~800)) 42.9
SABL Faster R-CNN (R-101-FPN, 1x, MS train=N) 41.7
SABL Cascade R-CNN (R-50-FPN, 1x, MS train=N) 41.6
SABL RetinaNet (R-101-FPN, 1x, GN=Y, MS train=N) 40.5
SABL Faster R-CNN (R-50-FPN, 1x, MS train=N) 39.9
SABL RetinaNet (R-101-FPN, 1x, GN=N, MS train=N) 39.7
SABL RetinaNet (R-50-FPN, 1x, GN=Y, MS train=N) 38.8
SABL RetinaNet (R-50-FPN, 1x, GN=N, MS train=N) 37.7
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