[NeurIPS 2021] SSUL: Semantic Segmentation with Unknown Label for Exemplar-based Class-Incremental Learning

Related tags

Deep LearningSSUL
Overview

SSUL - Official Pytorch Implementation (NeurIPS 2021)

SSUL: Semantic Segmentation with Unknown Label for Exemplar-based Class-Incremental Learning
Sungmin Cha1,2*, Beomyoung Kim3*, YoungJoon Yoo2,3, Taesup Moon1
* Equal contribution

1 Department of Electrical and Computer Engineering, Seoul National University
2 NAVER AI Lab
3 Face, NAVER Clova

NeurIPS 2021

Paper

Abtract

This paper introduces a solid state-of-the-art baseline for a class-incremental semantic segmentation (CISS) problem. While the recent CISS algorithms utilize variants of the knowledge distillation (KD) technique to tackle the problem, they failed to fully address the critical challenges in CISS causing the catastrophic forgetting; the semantic drift of the background class and the multi-label prediction issue. To better address these challenges, we propose a new method, dubbed SSUL-M (Semantic Segmentation with Unknown Label with Memory), by carefully combining techniques tailored for semantic segmentation. Specifically, we claim three main contributions. (1) defining unknown classes within the background class to help to learn future classes (help plasticity), (2) freezing backbone network and past classifiers with binary cross-entropy loss and pseudo-labeling to overcome catastrophic forgetting (help stability), and (3) utilizing tiny exemplar memory for the first time in CISS to improve both plasticity and stability. The extensively conducted experiments show the effectiveness of our method, achieving significantly better performance than the recent state-of-the-art baselines on the standard benchmark datasets. Furthermore, we justify our contributions with thorough ablation analyses and discuss different natures of the CISS problem compared to the traditional class-incremental learning targeting classification.

Experimental Results (mIoU all)

Method VOC 10-1 (11 tasks) VOC 15-1 (6 tasks) VOC 5-3 (6 tasks) VOC 19-1 (2 tasks) VOC 15-5 (2 tasks) VOC 5-1 (16 tasks) VOC 2-1 (19 tasks)
MiB 12.65 29.29 46.71 69.15 70.08 10.03 9.88
PLOP 30.45 54.64 18.68 73.54 70.09 6.46 4.47
SSUL 59.25 67.61 56.89 75.44 71.22 48.65 38.32
SSUL-M 64.12 71.37 58.37 76.49 73.02 55.11 44.74
Method ADE 100-5 (11 tasks) ADE 100-10 (6 tasks) ADE 100-50 (2 tasks) ADE 50-50 (3 tasks)
MiB 25.96 29.24 32.79 29.31
PLOP 28.75 31.59 32.94 30.40
SSUL 32.48 33.10 33.58 29.56
SSUL-M 34.56 34.46 34.37 29.77

Getting Started

Requirements

  • torch>=1.7.1
  • torchvision>=0.8.2
  • numpy
  • pillow
  • scikit-learn
  • tqdm
  • matplotlib

Datasets

data_root/
    --- VOC2012/
        --- Annotations/
        --- ImageSet/
        --- JPEGImages/
        --- SegmentationClassAug/
        --- saliency_map/
    --- ADEChallengeData2016
        --- annotations
            --- training
            --- validation
        --- images
            --- training
            --- validation

Download SegmentationClassAug and saliency_map

Class-Incremental Segmentation Segmentation on VOC 2012

DATA_ROOT=your_dataset_root_path
DATASET=voc
TASK=15-1 # [15-1, 10-1, 19-1, 15-5, 5-3, 5-1, 2-1, 2-2]
EPOCH=50
BATCH=32
LOSS=bce_loss
LR=0.01
THRESH=0.7
MEMORY=100 # [0 (for SSUL), 100 (for SSUL-M)]

python main.py --data_root ${DATA_ROOT} --model deeplabv3_resnet101 --gpu_id 0,1 --crop_val --lr ${LR} --batch_size ${BATCH} --train_epoch ${EPOCH} --loss_type ${LOSS} --dataset ${DATASET} --task ${TASK} --overlap --lr_policy poly --pseudo --pseudo_thresh ${THRESH} --freeze --bn_freeze --unknown --w_transfer --amp --mem_size ${MEMORY}

Class-Incremental Segmentation Segmentation on ADE20K

DATA_ROOT=your_dataset_root_path
DATASET=ade
TASK=100-5 # [100-5, 100-10, 100-50, 50-50]
EPOCH=100
BATCH=24
LOSS=bce_loss
LR=0.05
THRESH=0.7
MEMORY=300 # [0 (for SSUL), 300 (for SSUL-M)]

python main.py --data_root ${DATA_ROOT} --model deeplabv3_resnet101 --gpu_id 0,1 --crop_val --lr ${LR} --batch_size ${BATCH} --train_epoch ${EPOCH} --loss_type ${LOSS} --dataset ${DATASET} --task ${TASK} --overlap --lr_policy warm_poly --pseudo --pseudo_thresh ${THRESH} --freeze --bn_freeze --unknown --w_transfer --amp --mem_size ${MEMORY}

Qualitative Results

Acknowledgement

Our implementation is based on these repositories: DeepLabV3Plus-Pytorch, Torchvision.

License

SSUL
Copyright 2021-present NAVER Corp.

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in
all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
THE SOFTWARE.
Owner
Clova AI Research
Open source repository of Clova AI Research, NAVER & LINE
Clova AI Research
Pytorch implementation for "Large-Scale Long-Tailed Recognition in an Open World" (CVPR 2019 ORAL)

Large-Scale Long-Tailed Recognition in an Open World [Project] [Paper] [Blog] Overview Open Long-Tailed Recognition (OLTR) is the author's re-implemen

Zhongqi Miao 761 Dec 26, 2022
Python scripts for performing lane detection using the LSTR model in ONNX

ONNX LSTR Lane Detection Python scripts for performing lane detection using the Lane Shape Prediction with Transformers (LSTR) model in ONNX. Requirem

Ibai Gorordo 29 Aug 30, 2022
Text completion with Hugging Face and TensorFlow.js running on Node.js

Katana ML Text Completion 🤗 Description Runs with with Hugging Face DistilBERT and TensorFlow.js on Node.js distilbert-model - converter from Hugging

Katana ML 2 Nov 04, 2022
a dnn ai project to classify which food people are eating on audio recordings

Deep Learning - EAT Challenge About This project is part of an AI challenge of the DeepLearning course 2021 at the University of Augsburg. The objecti

Marco Tröster 1 Oct 24, 2021
Official PyTorch Implementation for "Recurrent Video Deblurring with Blur-Invariant Motion Estimation and Pixel Volumes"

PVDNet: Recurrent Video Deblurring with Blur-Invariant Motion Estimation and Pixel Volumes This repository contains the official PyTorch implementatio

Junyong Lee 98 Nov 06, 2022
A toy compiler that can convert Python scripts to pickle bytecode 🥒

Pickora 🐰 A small compiler that can convert Python scripts to pickle bytecode. Requirements Python 3.8+ No third-party modules are required. Usage us

ꌗᖘ꒒ꀤ꓄꒒ꀤꈤꍟ 68 Jan 04, 2023
MoveNetを用いたPythonでの姿勢推定のデモ

MoveNet-Python-Example MoveNetのPythonでの動作サンプルです。 ONNXに変換したモデルも同梱しています。変換自体を試したい方はMoveNet_tf2onnx.ipynbを使用ください。 2021/08/24時点でTensorFlow Hubで提供されている以下モデ

KazuhitoTakahashi 38 Dec 17, 2022
BaseCls BaseCls 是一个基于 MegEngine 的预训练模型库,帮助大家挑选或训练出更适合自己科研或者业务的模型结构

BaseCls BaseCls 是一个基于 MegEngine 的预训练模型库,帮助大家挑选或训练出更适合自己科研或者业务的模型结构。 文档地址:https://basecls.readthedocs.io 安装 安装环境 BaseCls 需要 Python = 3.6。 BaseCls 依赖 M

MEGVII Research 28 Dec 23, 2022
Code for "Neural Body: Implicit Neural Representations with Structured Latent Codes for Novel View Synthesis of Dynamic Humans" CVPR 2021 best paper candidate

News 05/17/2021 To make the comparison on ZJU-MoCap easier, we save quantitative and qualitative results of other methods at here, including Neural Vo

ZJU3DV 748 Jan 07, 2023
Housing Price Prediction

This project aim was to predict the price of houses in the Boston area during the great financial crisis through regression, as well as classify houses into different quality categories according to

Florian Klement 1 Jan 27, 2022
Pytorch implementation of YOLOX、PPYOLO、PPYOLOv2、FCOS an so on.

简体中文 | English miemiedetection 概述 miemiedetection是女装大佬咩酱基于YOLOX进行二次开发的个人检测库(使用的深度学习框架为pytorch),支持Windows、Linux系统,以女装大佬咩酱的名字命名。miemiedetection是一个不需要安装的

248 Jan 02, 2023
Official repository for the paper "GN-Transformer: Fusing AST and Source Code information in Graph Networks".

GN-Transformer AST This is the official repository for the paper "GN-Transformer: Fusing AST and Source Code information in Graph Networks". Data Prep

Cheng Jun-Yan 10 Nov 26, 2022
FusionNet: A deep fully residual convolutional neural network for image segmentation in connectomics

FusionNet_Pytorch FusionNet: A deep fully residual convolutional neural network for image segmentation in connectomics Requirements Pytorch 0.1.11 Pyt

Choi Gunho 102 Dec 13, 2022
Diffusion Normalizing Flow (DiffFlow) Neurips2021

Diffusion Normalizing Flow (DiffFlow) Reproduce setup environment The repo heavily depends on jam, a personal toolbox developed by Qsh.zh. The API may

76 Jan 01, 2023
Rotary Transformer

[中文|English] Rotary Transformer Rotary Transformer is an MLM pre-trained language model with rotary position embedding (RoPE). The RoPE is a relative

325 Jan 03, 2023
Deep learning model, heat map, data prepo

deep learning model, heat map, data prepo

Pamela Dekas 1 Jan 14, 2022
Video Frame Interpolation with Transformer (CVPR2022)

VFIformer Official PyTorch implementation of our CVPR2022 paper Video Frame Interpolation with Transformer Dependencies python = 3.8 pytorch = 1.8.0

DV Lab 63 Dec 16, 2022
Code for ICLR2018 paper: Improving GAN Training via Binarized Representation Entropy (BRE) Regularization - Y. Cao · W Ding · Y.C. Lui · R. Huang

code for "Improving GAN Training via Binarized Representation Entropy (BRE) Regularization" (ICLR2018 paper) paper: https://arxiv.org/abs/1805.03644 G

21 Oct 12, 2020
A PyTorch library and evaluation platform for end-to-end compression research

CompressAI CompressAI (compress-ay) is a PyTorch library and evaluation platform for end-to-end compression research. CompressAI currently provides: c

InterDigital 680 Jan 06, 2023
Light-Head R-CNN

Light-head R-CNN Introduction We release code for Light-Head R-CNN. This is my best practice for my research. This repo is organized as follows: light

jemmy li 835 Dec 06, 2022