Method Details

Details for method 'Pixelwise Instance Segmentation with a Dynamically Instantiated Network'


Method overview

name Pixelwise Instance Segmentation with a Dynamically Instantiated Network
challenge instance-level semantic labeling
details We propose an Instance Segmentation system that produces a segmentation map where each pixel is assigned an object class and instance identity label. Our method is based on an initial semantic segmentation module which feeds into an instance subnetwork. This subnetwork uses the initial category-level segmentation, along with cues from the output of an object detector, within an end-to-end CRF to predict instances. This part of our model is dynamically instantiated to produce a variable number of instances per image. Our end-to-end approach requires no post-processing and considers the image holistically, instead of processing independent proposals. As a result, it reasons about occlusions (unlike some related work, a single pixel cannot belong to multiple instances).
publication Pixelwise Instance Segmentation with a Dynamically Instantiated Network
Anurag Arnab and Philip Torr
Computer Vision and Pattern Recognition (CVPR) 2017
project page / code
used Cityscapes data fine annotations, coarse annotations
used external data ImageNet
runtime n/a
subsampling no
submission date January, 2017
previous submissions


Average results

Metric Value
AP 19.9906
AP50% 38.8278
AP100m 32.6009
AP50m 37.6238


Class results

Class AP AP50% AP100m AP50m
person 16.4671 37.117 31.8394 32.8176
rider 16.6669 42.0743 28.2143 29.0221
car 25.6754 45.7449 41.1294 44.1784
truck 20.5676 30.2345 33.6878 39.8547
bus 29.9921 44.6628 48.9322 60.6816
train 23.3616 40.5202 35.3079 51.6885
motorcycle 17.1229 41.838 23.9906 24.9035
bicycle 10.0715 28.4305 17.7052 17.8442



Download results as .csv file

Benchmark page