MOVE: Unsupervised Movable Object Segmentation and Detection (2022-10-14T00:00:00.000000Z)

Abstract

We introduce MOVE, a novel method to segment objects without any form of supervision. MOVE exploits the fact that foreground objects can be shifted locally relative to their initial position and result in realistic (undistorted) new images. This property allows us to train a segmentation model on a dataset of images without annotation and to achieve state of the art (SotA) performance on several evaluation datasets for unsupervised salient object detection and segmentation. In unsupervised single object discovery, MOVE gives an average CorLoc improvement of 7.2% over the SotA, and in unsupervised class-agnostic object detection it gives a relative AP improvement of 53% on average. Our approach is built on top of self-supervised features (e.g. from DINO or MAE), an inpainting network (based on the Masked AutoEncoder) and adversarial training.

Authors

P. Favaro

3 papers

Adam Bielski

2 papers

Abstract

Authors

References96 items

Board

Deep Spectral Methods: A Surprisingly Strong Baseline for Unsupervised Semantic Segmentation and Localization

Self-Supervised Learning of Object Parts for Semantic Segmentation

Towards Open-Set Object Detection and Discovery

Unsupervised Salient Object Detection with Spectral Cluster Voting

FreeSOLO: Learning to Segment Objects without Annotations

Self-Supervised Transformers for Unsupervised Object Discovery using Normalized Cut

Deep ViT Features as Dense Visual Descriptors

TransFGU: A Top-down Approach to Fine-Grained Unsupervised Semantic Segmentation

GANSeg: Learning to Segment by Unsupervised Hierarchical Image Generation

Masked Autoencoders Are Scalable Vision Learners

Projected GANs Converge Faster

Self-supervised Human Detection and Segmentation via Background Inpainting

Localizing Objects with Self-Supervised Transformers and no Labels

Per-Pixel Classification is Not All You Need for Semantic Segmentation

Large-Scale Unsupervised Object Discovery

DETReg: Unsupervised Pretraining with Region Priors for Object Detection

Finding an Unsupervised Image Segmenter in Each of Your Deep Generative Models

The Pursuit of Knowledge: Discovering and Localizing Novel Categories using Dual Memory

Emerging Properties in Self-Supervised Vision Transformers

An Empirical Study of Training Self-Supervised Vision Transformers

Information-Theoretic Segmentation by Inpainting Error Maximization

Dense Contrastive Learning for Self-Supervised Visual Pre-Training

An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale

Autoregressive Unsupervised Image Segmentation

Toward unsupervised, multi-object discovery in large-scale image collections

Unsupervised Learning of Visual Features by Contrasting Cluster Assignments

Object Segmentation Without Labels with Large-Scale Generative Models

End-to-End Object Detection with Transformers

Agriculture-Vision: A Large Aerial Image Database for Agricultural Pattern Analysis

OneGAN: Simultaneous Unsupervised Learning of Conditional Image Generation, Foreground Segmentation, and Fine-Grained Clustering

PyTorch: An Imperative Style, High-Performance Deep Learning Library

Momentum Contrast for Unsupervised Visual Representation Learning

Contrastive Multiview Coding

Unsupervised Object Segmentation by Redrawing

Object Discovery with a Copy-Pasting GAN

Emergence of Object Segmentation in Perturbed Generative Models

Objects as Points

Unsupervised Image Matching and Object Discovery as Optimization

Unsupervised object discovery and co-localization by deep descriptor transformation

Object Discovery From a Single Unlabeled Image by Mining Frequent Itemsets With Multi-Scale Features

MONet: Unsupervised Scene Decomposition and Representation

Unsupervised Moving Object Detection via Contextual Information Separation

SEIGAN: Towards Compositional Image Generation by Simultaneously Learning to Segment, Enhance, and Inpaint

Investigating object compositionality in Generative Adversarial Networks

Invariant Information Clustering for Unsupervised Image Classification and Segmentation

Unsupervised Image Segmentation by Backpropagation

Deep Unsupervised Saliency Detection: A Multiple Noisy Labeling Perspective

Learning to Segment via Cut-and-Paste

Learning to Segment Every Thing

W-Net: A Deep Model for Fully Unsupervised Image Segmentation

Learning to Detect Salient Objects with Image-Level Supervision

Mask R-CNN

LR-GAN: Layered Recursive Generative Adversarial Networks for Image Generation

Generating Images Part by Part with Composite Generative Adversarial Networks

Hierarchical Image Saliency Detection on Extended CSSD

Attend, Infer, Repeat: Fast Scene Understanding with Generative Models

Simple Does It: Weakly Supervised Instance and Semantic Segmentation

The Fast Bilateral Solver

A weighted sparse coding framework for saliency detection

DeepDriving: Learning Affordance for Direct Perception in Autonomous Driving

Visual saliency based on multiscale deep features

Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift

Medical image segmentation on GPUs - A comprehensive review

Unsupervised object discovery and localization in the wild: Part-based matching with bottom-up region proposals

Adam: A Method for Stochastic Optimization

Edge Boxes: Locating Object Proposals from Edges

Saliency Optimization from Robust Background Detection

Microsoft COCO: Common Objects in Context

Hierarchical Saliency Detection

Saliency Detection via Graph-Based Manifold Ranking

Looking Beyond the Image: Unsupervised Learning for Object Saliency and Detection

Selective Search for Object Recognition

The Caltech-UCSD Birds-200-2011 Dataset

Localizing Objects While Learning Their Appearance

Unsupervised Detection of Regions of Interest Using Iterative Link Analysis

ImageNet: A large-scale hierarchical image database