Published on Tue Dec 04 2018

Learning to Fuse Things and Stuff

Jie Li, Allan Raventos, Arjun Bhargava, Takaaki Tagawa, Adrien Gaidon

We propose an end-to-end learning approach for panoptic segmentation. Our model uses feature maps from a shared backbone network to predict in a single feed-forward pass both things and stuff segmentations.

0
0
0
Abstract

We propose an end-to-end learning approach for panoptic segmentation, a novel task unifying instance (things) and semantic (stuff) segmentation. Our model, TASCNet, uses feature maps from a shared backbone network to predict in a single feed-forward pass both things and stuff segmentations. We explicitly constrain these two output distributions through a global things and stuff binary mask to enforce cross-task consistency. Our proposed unified network is competitive with the state of the art on several benchmarks for panoptic segmentation as well as on the individual semantic and instance segmentation tasks.

Tue Jan 08 2019
Computer Vision
Panoptic Feature Pyramid Networks
The recently introduced panoptic segmentation task has renewed our community's interest in unifying the tasks of instance segmentation and semantic segmentation. Current methods for this joint task use separate and dissimilar networks. We aim to unify these methods at the architectural level, designing a single
0
0
0
Tue Jan 14 2020
Computer Vision
Unifying Training and Inference for Panoptic Segmentation
We present an end-to-end network to bridge the gap between training andference pipeline for panoptic segmentation. The task seeks to partition an image into semantic regions for "stuff" and object instances for "things"
0
0
0
Thu Sep 06 2018
Computer Vision
Panoptic Segmentation with a Joint Semantic and Instance Segmentation Network
We present a single network method for panoptic segmentation. This method combines the predictions from a jointly trained semantic and instance segmentation network using heuristics. Results for this method are submitted to the COCO and Mapillary Joint Recognition Challenge.
0
0
0
Sat Jan 12 2019
Computer Vision
UPSNet: A Unified Panoptic Segmentation Network
The UPSNet achieves state-of-the-art performance with much faster inference. Code has been made available at: https://github.com/uber-research/UPSNet.
0
0
0
Thu Oct 10 2019
Machine Learning
Panoptic-DeepLab
We present Panoptic-DeepLab, a bottom-up and single-shot approach for panoptic segmentation. Our single Panopti-Deep lab sets the new state-of-art at all three Cityscapes benchmarks.
0
0
0
Wed Oct 09 2019
Computer Vision
Fast Panoptic Segmentation Network
Fast Panoptic Segmentation Network (FPSNet) does not require computationally costly instance mask predictions or merging heuristics. FPSNet is faster than existing panoptic segmentation methods, while achieving better or similar performance.
0
0
0