Showing 1–1 of 1 results for author: Prisadnikov, N

A Simple and Generalist Approach for Panoptic Segmentation

Authors: Nedyalko Prisadnikov, Wouter Van Gansbeke, Danda Pani Paudel, Luc Van Gool

Abstract: Panoptic segmentation is an important computer vision task, where the current state-of-the-art solutions require specialized components to perform well. We propose a simple generalist framework based on a deep encoder - shallow decoder architecture with per-pixel prediction. Essentially fine-tuning a massively pretrained image model with minimal additional components. Naively this method does not… ▽ More Panoptic segmentation is an important computer vision task, where the current state-of-the-art solutions require specialized components to perform well. We propose a simple generalist framework based on a deep encoder - shallow decoder architecture with per-pixel prediction. Essentially fine-tuning a massively pretrained image model with minimal additional components. Naively this method does not yield good results. We show that this is due to imbalance during training and propose a novel method for reducing it - centroid regression in the space of spectral positional embeddings. Our method achieves panoptic quality (PQ) of 55.1 on the challenging MS-COCO dataset, state-of-the-art performance among generalist methods. △ Less

Submitted 7 March, 2025; v1 submitted 29 August, 2024; originally announced August 2024.