TASK	DATASET	MODEL	METRIC NAME	METRIC VALUE	GLOBAL RANK
Real-Time Object Detection	COCO	SpineNet-49 (RetinaNet, single-scale, with TRT)	MAP	46.7	# 4
Real-Time Object Detection	COCO	SpineNet-49 (RetinaNet, single-scale, with TRT)	FPS	29	# 10
Real-Time Object Detection	COCO	SpineNet-49 (RetinaNet, single-scale, with TRT)	inference time (ms)	34.3	# 2
Instance Segmentation	COCO minival	SpineNet-190 (Mask R-CNN, single-scale)	mask AP	46.1	# 2
Instance Segmentation	COCO minival	SpineNet-190 (Mask R-CNN, single-scale)	AP50	70.6	# 1
Instance Segmentation	COCO minival	SpineNet-190 (Mask R-CNN, single-scale)	AP75	50.1	# 1
Instance Segmentation	COCO minival	SpineNet-190 (Mask R-CNN, single-scale)	APL	63.6	# 1
Instance Segmentation	COCO minival	SpineNet-190 (Mask R-CNN, single-scale)	APM	49.2	# 1
Instance Segmentation	COCO minival	SpineNet-190 (Mask R-CNN, single-scale)	APS	27.5	# 1
Object Detection	COCO minival	SpineNet-190 (RetinaNet, single-scale)	box AP	52.0	# 5
Object Detection	COCO minival	SpineNet-190 (RetinaNet, single-scale)	AP50	71.5	# 3
Object Detection	COCO minival	SpineNet-190 (RetinaNet, single-scale)	AP75	56.5	# 4
Object Detection	COCO minival	SpineNet-190 (RetinaNet, single-scale)	APS	37.4	# 1
Object Detection	COCO minival	SpineNet-190 (RetinaNet, single-scale)	APM	55.6	# 3
Object Detection	COCO minival	SpineNet-190 (RetinaNet, single-scale)	APL	65.0	# 4
Object Detection	COCO minival	SpineNet-190 (Mask R-CNN, single-scale)	box AP	52.2	# 4
Object Detection	COCO minival	SpineNet-190 (Mask R-CNN, single-scale)	AP50	72.6	# 1
Object Detection	COCO minival	SpineNet-190 (Mask R-CNN, single-scale)	AP75	57.2	# 2
Object Detection	COCO minival	SpineNet-190 (Mask R-CNN, single-scale)	APS	33.9	# 4
Object Detection	COCO minival	SpineNet-190 (Mask R-CNN, single-scale)	APM	54.7	# 4
Object Detection	COCO minival	SpineNet-190 (Mask R-CNN, single-scale)	APL	68.8	# 1
Instance Segmentation	COCO test-dev	SpineNet-190 (Mask R-CNN, single-scale)	mask AP	46.3	# 2
Instance Segmentation	COCO test-dev	SpineNet-190 (Mask R-CNN, single-scale)	AP50	71.0	# 2
Instance Segmentation	COCO test-dev	SpineNet-190 (Mask R-CNN, single-scale)	AP75	50.5	# 3
Instance Segmentation	COCO test-dev	SpineNet-190 (Mask R-CNN, single-scale)	APS	29.7	# 3
Instance Segmentation	COCO test-dev	SpineNet-190 (Mask R-CNN, single-scale)	APM	49.0	# 3
Instance Segmentation	COCO test-dev	SpineNet-190 (Mask R-CNN, single-scale)	APL	58.5	# 5
Object Detection	COCO test-dev	SpineNet-190 (RetinaNet, single-scale)	box AP	52.1	# 6
Object Detection	COCO test-dev	SpineNet-190 (RetinaNet, single-scale)	AP50	71.8	# 6
Object Detection	COCO test-dev	SpineNet-190 (RetinaNet, single-scale)	AP75	56.5	# 7
Object Detection	COCO test-dev	SpineNet-190 (RetinaNet, single-scale)	APS	35.4	# 5
Object Detection	COCO test-dev	SpineNet-190 (RetinaNet, single-scale)	APM	55.0	# 7
Object Detection	COCO test-dev	SpineNet-190 (RetinaNet, single-scale)	APL	63.6	# 11
Object Detection	COCO test-dev	SpineNet-190 (Mask R-CNN, single-scale)	box AP	52.5	# 5
Object Detection	COCO test-dev	SpineNet-190 (Mask R-CNN, single-scale)	AP50	73.2	# 2
Object Detection	COCO test-dev	SpineNet-190 (Mask R-CNN, single-scale)	AP75	57.8	# 5
Object Detection	COCO test-dev	SpineNet-190 (Mask R-CNN, single-scale)	APS	36.0	# 2
Object Detection	COCO test-dev	SpineNet-190 (Mask R-CNN, single-scale)	APM	54.6	# 9
Object Detection	COCO test-dev	SpineNet-190 (Mask R-CNN, single-scale)	APL	63.9	# 9

Badge	Markdown
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/spinenet-learning-scale-permuted-backbone-for/instance-segmentation-on-coco-minival)](https://paperswithcode.com/sota/instance-segmentation-on-coco-minival?p=spinenet-learning-scale-permuted-backbone-for)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/spinenet-learning-scale-permuted-backbone-for/instance-segmentation-on-coco)](https://paperswithcode.com/sota/instance-segmentation-on-coco?p=spinenet-learning-scale-permuted-backbone-for)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/spinenet-learning-scale-permuted-backbone-for/real-time-object-detection-on-coco)](https://paperswithcode.com/sota/real-time-object-detection-on-coco?p=spinenet-learning-scale-permuted-backbone-for)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/spinenet-learning-scale-permuted-backbone-for/object-detection-on-coco-minival)](https://paperswithcode.com/sota/object-detection-on-coco-minival?p=spinenet-learning-scale-permuted-backbone-for)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/spinenet-learning-scale-permuted-backbone-for/object-detection-on-coco)](https://paperswithcode.com/sota/object-detection-on-coco?p=spinenet-learning-scale-permuted-backbone-for)`

SpineNet: Learning Scale-Permuted Backbone for Recognition and Localization

CVPR 2020 • Xianzhi Du • Tsung-Yi Lin • Pengchong Jin • Golnaz Ghiasi • Mingxing Tan • Yin Cui • Quoc V. Le • Xiaodan Song

Convolutional neural networks typically encode an input image into a series of intermediate features with decreasing resolutions. While this structure is suited to classification tasks, it does not perform well for tasks requiring simultaneous recognition and localization (e.g., object detection)... (read more)

PDF Abstract

Code

Add Remove

tensorflow/tpu

3,647

lucifer443/SpineNet-Pytorch

Tasks

Add Remove

INSTANCE SEGMENTATION

NEURAL ARCHITECTURE SEARCH

OBJECT DETECTION

REAL-TIME OBJECT DETECTION

Results from the Paper

Edit

#2 best model for Instance Segmentation on COCO minival

Get a GitHub badge

TASK	DATASET	MODEL	METRIC NAME	METRIC VALUE	GLOBAL RANK	LEADERBOARD
Real-Time Object Detection	COCO	SpineNet-49 (RetinaNet, single-scale, with TRT)	MAP	46.7	# 4	Compare
			FPS	29	# 10	Compare
			inference time (ms)	34.3	# 2	Compare
Instance Segmentation	COCO minival	SpineNet-190 (Mask R-CNN, single-scale)	mask AP	46.1	# 2	Compare
			AP50	70.6	# 1	Compare
			AP75	50.1	# 1	Compare
			APL	63.6	# 1	Compare
			APM	49.2	# 1	Compare
			APS	27.5	# 1	Compare
Object Detection	COCO minival	SpineNet-190 (RetinaNet, single-scale)	box AP	52.0	# 5	Compare
			AP50	71.5	# 3	Compare
			AP75	56.5	# 4	Compare
			APS	37.4	# 1	Compare
			APM	55.6	# 3	Compare
			APL	65.0	# 4	Compare
Object Detection	COCO minival	SpineNet-190 (Mask R-CNN, single-scale)	box AP	52.2	# 4	Compare
			AP50	72.6	# 1	Compare
			AP75	57.2	# 2	Compare
			APS	33.9	# 4	Compare
			APM	54.7	# 4	Compare
			APL	68.8	# 1	Compare
Instance Segmentation	COCO test-dev	SpineNet-190 (Mask R-CNN, single-scale)	mask AP	46.3	# 2	Compare
			AP50	71.0	# 2	Compare
			AP75	50.5	# 3	Compare
			APS	29.7	# 3	Compare
			APM	49.0	# 3	Compare
			APL	58.5	# 5	Compare
Object Detection	COCO test-dev	SpineNet-190 (RetinaNet, single-scale)	box AP	52.1	# 6	Compare
			AP50	71.8	# 6	Compare
			AP75	56.5	# 7	Compare
			APS	35.4	# 5	Compare
			APM	55.0	# 7	Compare
			APL	63.6	# 11	Compare
Object Detection	COCO test-dev	SpineNet-190 (Mask R-CNN, single-scale)	box AP	52.5	# 5	Compare
			AP50	73.2	# 2	Compare
			AP75	57.8	# 5	Compare
			APS	36.0	# 2	Compare
			APM	54.6	# 9	Compare
			APL	63.9	# 9	Compare

Methods used in the Paper

Add Remove

METHOD	TYPE
SpineNet	Convolutional Neural Networks
ReLU	Activation Functions
Swish	Activation Functions
Linear Warmup With Cosine Annealing	Learning Rate Schedules
Batch Normalization	Normalization
SGD with Momentum	Stochastic Optimization
Weight Decay	Regularization
Stochastic Depth	Regularization
Softmax	Output Functions
Global Average Pooling	Pooling Operations
Dense Connections	Feedforward Networks
Mask R-CNN	Instance Segmentation Models
RoIAlign	RoI Feature Extractors
Convolution	Convolutions
RetinaNet	Object Detection Models
FPN	Feature Extractors
Focal Loss	Loss Functions
ResNet	Convolutional Neural Networks
Random Resized Crop	Image Data Augmentation
Random Horizontal Flip	Image Data Augmentation
Max Pooling	Pooling Operations
1x1 Convolution	Convolutions
Kaiming Initialization	Initialization
Residual Block	Skip Connection Blocks
Bottleneck Residual Block	Skip Connection Blocks
LSTM	Recurrent Neural Networks
Sigmoid Activation	Activation Functions
Average Pooling	Pooling Operations
Residual Connection	Skip Connections
Tanh Activation	Activation Functions

SpineNet: Learning Scale-Permuted Backbone for Recognition and Localization

Code Edit Add Remove

Tasks Edit Add Remove

Results from the Paper Edit

Methods used in the Paper Edit Add Remove

Code

Add Remove

Tasks

Add Remove

Results from the Paper

Edit

Methods used in the Paper

Add Remove