Frugal learning

Collecting diverse enough data, and annotating it precisely, is complex, costly and time-consuming. To reduce dramatically these needs, we explore various alternatives to fully-supervised learning, e.g, training that is unsupervised (as rOSD at ECCCV’20), self-supervised (as BoWNet at CVPR’20), semi-supervised, active, zero-shot (as ZS3 at NeurIPS’19) or few-shot. We also investigate training with fully-synthetic data (in combination with unsupervised domain adaptation) and with GAN-augmented data (as Semantic Palette at CVPR’21).

Frugal learning

Publications

Three Pillars improving Vision Foundation Model Distillation for Lidar

SPOT: Self-Training with Patch-Order Permutation for Object-Centric Learning with Autoregressive Transformers

A Simple Recipe for Language-guided Domain Generalized Segmentation

Unsupervised Object Localization in the Era of Self-Supervised ViTs: A Survey

SALUDA: Surface-based Automotive Lidar Unsupervised Domain Adaptation

MOCA: Self-supervised Representation Learning by Predicting Masked Online Codebook Assignments

BEVContrast: Self-Supervision in BEV Space for Automotive Lidar Point Clouds

CLIP-DIY: CLIP Dense Inference Yields Open-Vocabulary Semantic Segmentation For-Free

POP-3D: Open-Vocabulary 3D Occupancy Prediction from Images

You Never Get a Second Chance To Make a Good First Impression: Seeding Active Learning for 3D Semantic Segmentation

PØDA: Prompt-driven Zero-shot Domain Adaptation

Unsupervised Object Localization: Observing the Background to Discover Objects

RangeViT: Towards Vision Transformers for 3D Semantic Segmentation in Autonomous Driving

ALSO: Automotive Lidar Self-supervision by Occupancy estimation

What to Hide from Your Students: Attention-Guided Masked Image Modeling

Drive&Segment: Unsupervised Semantic Segmentation of Urban Scenes via Cross-modal Distillation

Active Learning Strategies for Weakly-Supervised Object Detection

CSG0: Continual Urban Scene Generation with Zero Forgetting

Image-to-Lidar Self-Supervised Distillation for Autonomous Driving Data

Localizing Objects with Self-Supervised Transformers and no Labels

Large-Scale Unsupervised Object Discovery

Handling new target classes in semantic segmentation with domain adaptation

Generative Zero-Shot Learning for Semantic Segmentation of 3D Point Clouds

Online Bag-of-Visual-Words Generation for Unsupervised Representation Learning

Semantic Palette: Guiding Scene Generation with Class Proportions

Artificial Dummies for Urban Dataset Augmentation

Toward Unsupervised, Multi-Object Discovery in Large-Scale Image Collections

Learning Representations by Predicting Bags of Visual Words

This dataset does not exist: training models from generated images

Zero-Shot Semantic Segmentation

Boosting Few-Shot Visual Learning With Self-Supervision

Unsupervised Image Matching and Object Discovery as Optimization