A2LP for short, ECCV2020 spotlight, Investigating SSL principles for UDA problems

Last update: Oct 27, 2022

Overview

Label-Propagation-with-Augmented-Anchors (A2LP)

Official codes of the ECCV2020 spotlight (label propagation with augmented anchors: a simple semi-supervised learning baseline for unsupervised domain adaptation) [Paper]. In this work, we investigating SSL principles for UDA problems.

One-sentence Summary

Proper algorithmic adaptation should be made when applying the SSL techniques to UDA tasks, even both tasks of UDA and SSL adopt the labeled and unlabeled data as the input.

Usage

Please refer to the 'run.sh'. We also provide the corresponding log file in the file of './test'
You can start with these examples easily.  

NOTE: Results are based on the ImageNet pre-trained features !!! No additional training involved. 

Note that the A2LP can introduce excellent pseudo labels of unlabeled target data in DA 
(compared to the FC-based classifier and the clustering algorithm). Therefore it could
empower algorithms of DA using pseudo labels of unlabeled target data.

Requirement

PyTorch 1.2.0
spherecluster
nndescent

Dataset

The structure of the dataset should be like

Office-31
|_ amazon
|  |_ back_pack
|     |_ <im-1-name>.jpg
|     |_ ...
|     |_ <im-N-name>.jpg
|  |_ bike
|     |_ <im-1-name>.jpg
|     |_ ...
|     |_ <im-N-name>.jpg
|  |_ ...
|_ dslr
|  |_ back_pack
|     |_ <im-1-name>.jpg
|     |_ ...
|     |_ <im-N-name>.jpg
|  |_ bike
|     |_ <im-1-name>.jpg
|     |_ ...
|     |_ <im-N-name>.jpg
|  |_ ...
|_ ...

Citation

@inproceedings{zhang2020label,
  title={Label Propagation with Augmented Anchors: A Simple Semi-Supervised Learning baseline for Unsupervised Domain Adaptation},
  author={Zhang, Yabin and Deng, Bin and Jia, Kui and Zhang, Lei},
  booktitle={European Conference on Computer Vision},
  pages={781--797},
  year={2020},
  organization={Springer}
}

Contact

If you have any problem about our code, feel free to contact
- zha[email protected]

or describe your problem in Issues.

A2LP for short, ECCV2020 spotlight, Investigating SSL principles for UDA problems

Related tags

Overview

Label-Propagation-with-Augmented-Anchors (A2LP)

One-sentence Summary

Usage

Requirement

Dataset

Citation

Contact

Owner

CLUES: Few-Shot Learning Evaluation in Natural Language Understanding

PushForKiCad - AISLER Push for KiCad EDA

Bot developed in Python that automates races in pegaxy.

[CVPR'22] COAP: Learning Compositional Occupancy of People

Deep Implicit Moving Least-Squares Functions for 3D Reconstruction

PyTorch implementation of: Michieli U. and Zanuttigh P., "Continual Semantic Segmentation via Repulsion-Attraction of Sparse and Disentangled Latent Representations", CVPR 2021.

내가 보려고 정리한 <프로그래밍 기초 Ⅰ> / organized for me

TensorFlow implementation for Bayesian Modeling and Uncertainty Quantification for Learning to Optimize: What, Why, and How

ICCV2021 Papers with Code

Pyramid R-CNN: Towards Better Performance and Adaptability for 3D Object Detection

S-attack library. Official implementation of two papers "Are socially-aware trajectory prediction models really socially-aware?" and "Vehicle trajectory prediction works, but not everywhere".

Dense Deep Unfolding Network with 3D-CNN Prior for Snapshot Compressive Imaging, ICCV2021 [PyTorch Code]

Deep Learning to Create StepMania SM FIles

Official implementation of "A Unified Objective for Novel Class Discovery", ICCV2021 (Oral)

Einshape: DSL-based reshaping library for JAX and other frameworks.

ServiceX Transformer that converts flat ROOT ntuples into columnwise data

Denoising Diffusion Probabilistic Models

Harmonious Textual Layout Generation over Natural Images via Deep Aesthetics Learning

TilinGNN: Learning to Tile with Self-Supervised Graph Neural Network (SIGGRAPH 2020)

[ICLR 2021] "CPT: Efficient Deep Neural Network Training via Cyclic Precision" by Yonggan Fu, Han Guo, Meng Li, Xin Yang, Yining Ding, Vikas Chandra, Yingyan Lin