ReMaX：放鬆以改善高效泛域分割訓練

ReMaX: Relaxing for Better Training on Efficient Panoptic Segmentation

June 29, 2023

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

cs.AI

摘要

本文提出了一種新機制，以促進面向全景分割的遮罩變壓器的訓練，實現其部署的民主化。我們觀察到，由於其高複雜性，面向全景分割的訓練目標將不可避免地導致更高的偽陽性懲罰。這種不平衡的損失使得基於端對端遮罩變壓器架構的訓練過程變得困難，特別是對於高效模型。本文提出了ReMaX，它在面向全景分割的訓練過程中為遮罩預測和類別預測添加了放鬆。我們展示通過這些簡單的放鬆技術，在訓練過程中，我們的模型可以通過明顯的邊界持續改進，而無需額外的推斷計算成本。通過將我們的方法與MobileNetV3-Small等高效骨幹結合，我們的方法在COCO、ADE20K和Cityscapes上實現了高效全景分割的新最先進結果。代碼和預訓練檢查點將在https://github.com/google-research/deeplab2 提供。

English

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.