Simple Copy-Paste is a Strong Data Augmentation Method for Instance Segmentation

Building instance segmentation models that are data-efficient and can handle rare object categories is an important challenge in computer vision. Leveraging data augmentations is a promising direction towards addressing this challenge... (read more)

PDF Abstract

Results from the Paper


 Ranked #1 on Object Detection on COCO minival (using extra training data)

     Get a GitHub badge
TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK USES EXTRA
TRAINING DATA
RESULT BENCHMARK
Object Detection COCO minival Cascade Eff-B7 NAS-FPN (1280, self-training Copy Paste, single-scale) box AP 57.0 # 1
Instance Segmentation COCO minival Cascade Eff-B7 NAS-FPN (1280, self-training Copy Paste, single-scale) mask AP 48.9 # 1
Object Detection COCO test-dev Cascade Eff-B7 NAS-FPN (1280, self-training Copy Paste, single-scale) box AP 57.3 # 1
Instance Segmentation COCO test-dev Cascade Eff-B7 NAS-FPN (1280, self-training Copy Paste, single-scale) mask AP 49.1 # 1
Instance Segmentation LVIS v1.0 Eff-B7 NAS-FPN (1280, Copy-Paste pre-training)) mask AP 38.1 # 1
Object Detection LVIS v1.0 Eff-B7 NAS-FPN (1280, Copy-Paste pre-training)) box AP 41.6 # 1
Object Detection PASCAL VOC 2007 Cascade Eff-B7 NAS-FPN (Copy Paste pre-training, single-scale) MAP 89.1% # 1
Semantic Segmentation PASCAL VOC 2012 val Eff-B7 NAS-FPN (Copy-Paste pre-training, single-scale)) mIoU 86.6% # 2

Methods used in the Paper


METHOD TYPE
Pointwise Convolution
Convolutions
Depthwise Convolution
Convolutions
Depthwise Separable Convolution
Convolutions
Entropy Regularization
Regularization
Sigmoid Activation
Activation Functions
PPO
Policy Gradient Methods
Residual Connection
Skip Connections
Tanh Activation
Activation Functions
LSTM
Recurrent Neural Networks
Stochastic Depth
Regularization
Max Pooling
Pooling Operations
Bottleneck Residual Block
Skip Connection Blocks
Residual Block
Skip Connection Blocks
Kaiming Initialization
Initialization
ResNet
Convolutional Neural Networks
Softmax
Output Functions
Mask R-CNN
Instance Segmentation Models
RoIAlign
RoI Feature Extractors
Cascade Mask R-CNN
Instance Segmentation Models
Neural Architecture Search
Neural Architecture Search
Batch Normalization
Normalization
ReLU
Activation Functions
Global Average Pooling
Pooling Operations
NAS-FPN
Feature Extractors
RMSProp
Stochastic Optimization
Squeeze-and-Excitation Block
Image Model Blocks
Swish
Activation Functions
Dropout
Regularization
Average Pooling
Pooling Operations
Dense Connections
Feedforward Networks
Convolution
Convolutions
1x1 Convolution
Convolutions
Inverted Residual Block
Skip Connection Blocks
EfficientNet
Image Models
Copy-Paste
Image Data Augmentation