TensorX
返回文献探索

Paper · arXiv 2306.17319

ReMaX: Relaxing for Better Training on Efficient Panoptic Segmentation

Shuyang Sun, Weijun Wang, Qihang Yu, Andrew Howard, Philip Torr, Liang-Chieh Chen

3 upvotesJune 29, 2023arXiv 预印本
AI 摘要

ReMaX introduces relaxation techniques to improve mask transformer training for panoptic segmentation without extra inference costs, achieving state-of-the-art results with efficient backbones on benchmarks like COCO, ADE20K, and Cityscapes.

mask transformerspanoptic segmentationtraining objectivefalse positive penalizationunbalanced lossrelaxation techniquesMobileNetV3-SmallCOCOADE20KCityscapesDeeplab2

Abstract

This paper presents a new mechanism to facilitate the training of mask transformers for efficient panoptic segmentation, democratizing its deployment. We observe that due to its high complexity, the training objective of panoptic segmentation will inevitably lead to much higher false positive penalization. Such unbalanced loss makes the training process of the end-to-end mask-transformer based architectures difficult, especially for efficient models. In this paper, we present ReMaX that adds relaxation to mask predictions and class predictions during training for panoptic segmentation. We demonstrate that via these simple relaxation techniques during training, our model can be consistently improved by a clear margin without any extra computational cost on inference. By combining our method with efficient backbones like MobileNetV3-Small, our method achieves new state-of-the-art results for efficient panoptic segmentation on COCO, ADE20K and Cityscapes. Code and pre-trained checkpoints will be available at https://github.com/google-research/deeplab2.

北京市昌平区探索星信息技术及软件开发工作室

京ICP备2026059466号
ReMaX: Relaxing for Better Training on Efficient Panoptic Segmentation | TensorX