A novel Region of Interest Extraction Layer for Instance Segmentation

28 Apr 2020 Leonardo Rossi Akbar Karimi Andrea Prati

Given the wide diffusion of deep neural network architectures for computer vision tasks, several new applications are nowadays more and more feasible. Among them, a particular attention has been recently given to instance segmentation, by exploiting the results achievable by two-stage networks (such as Mask R-CNN or Faster R-CNN), derived from R-CNN... (read more)

PDF Abstract

Datasets


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK RESULT BENCHMARK
Instance Segmentation COCO minival GCnet (ResNet-50-FPN, GRoIE) mask AP 37.2 # 31
AP50 59.3 # 6
AP75 39.8 # 9
APL 51.2 # 6
APM 41 # 6
APS 20.2 # 6
Object Detection COCO minival Faster R-CNN (ResNet-50-FPN, GRoIE) box AP 37.5 # 102
AP50 59.2 # 58
AP75 40.6 # 65
APS 22.3 # 55
APM 41.5 # 56
APL 47.8 # 66
Instance Segmentation COCO minival Mask R-CNN (ResNet-50-FPN, GRoIE) mask AP 35.8 # 34
AP50 57.1 # 10
AP75 38.0 # 10
APL 48.7 # 7
APM 39 # 8
APS 19.1 # 7
Object Detection COCO minival Mask R-CNN (ResNet-50-FPN, GRoIE) box AP 38.4 # 95
AP50 59.9 # 51
AP75 41.7 # 58
APS 22.9 # 49
APM 42.1 # 53
APL 49.7 # 61

Methods used in the Paper


METHOD TYPE
Non-Local Operation
Image Feature Extractors
Residual Connection
Skip Connections
Non-Local Block
Image Model Blocks
GRoIE
RoI Feature Extractors
1x1 Convolution
Convolutions
FPN
Feature Extractors
Softmax
Output Functions
Convolution
Convolutions
RoIAlign
RoI Feature Extractors
Mask R-CNN
Instance Segmentation Models