Reviving Iterative Training with Mask Guidance for Interactive Segmentation | VitaVision
PaperarXiv (Cornell University) · 2021

Reviving Iterative Training with Mask Guidance for Interactive Segmentation

Konstantin SofiiukIlya A. PetrovAnton Konushin
arXiv 2102.06583DOIView in graph
1page built on it
1more page cites it
1narrative
0registry papers cite it

In the Atlas

ModelPrimary source of
RITM

Feedforward click-based interactive segmentation: HRNet+OCR encoder-decoder taking RGB + positive/negative disk-encoded clicks + previous mask, trained with iterative click simulation and Normalized Focal Loss on COCO+LVIS — sets a new state of the art without inference-time backward passes.

Cited as a reference by 1 page, grouped by domain

Segmentation 1
FocalClick

Appears in 1 narrative

NarrativeSegmentation, from Energy to Prompt

Lineage

Papers in the Atlas registry only · one dot per paper
Built upon by · 0
’21
Builds on · 0
Built upon by · 0