Segmentations(Instance & Panoptic)

Semantic segmentation(stuff + things) = pixel-wise classification

object detection

 

Instance Segmentation = Semantic segmentation + distinguishing instances

 

two-stage 

Mask R-CNN = Faster R-CNN + Mask branch

RoIAlign

 

single-stage

YOLACT(You Only Look At CoefficienTs) - real-time 연산은 어려움

mask는 아니지만 mask처럼 쓰이는 component인 prototypes(재료)를 구하여 span 하여(선형결합) mask를 만들고자 함

 

YolactEdge - (Yolact의 key 부분만 가져와서) real-time연산 가능하도록 노력

아직 연산이 smooth하진 않음

 

Panoptic segmentation(stuff + instances of Things)

 

UPSNet - FPN feature

Semantic + Instance head -> Panoptic head -> Panoptic logits

Instance head = find instance

Semantic head = Thing + stuff

 

VPSNet (for video)

1. Align refernece features onto the target feature map(Fusion at pixel level)

2. Track module associates different object instances(Track at instance level)

3. Fused-and-tracked modules are trained to synergize each other