https://github.com/universome/class-norm

Class Normalization for Continual Zero-Shot Learning
https://github.com/universome/class-norm

continual-learning iclr2021 initialization lifelong-learning normalization zero-shot-learning

Last synced: 6 days ago
JSON representation

Class Normalization for Continual Zero-Shot Learning

• Host: GitHub
• URL: https://github.com/universome/class-norm
• Owner: universome
• Created: 2019-10-02T10:02:43.000Z (over 4 years ago)
• Default Branch: master
• Last Pushed: 2021-02-05T11:37:36.000Z (over 3 years ago)
• Last Synced: 2024-04-18T14:42:54.876Z (2 months ago)
• Topics: continual-learning, iclr2021, initialization, lifelong-learning, normalization, zero-shot-learning
• Language: Python
• Homepage:
• Size: 70.9 MB
• Stars: 34
• Watchers: 5
• Forks: 3
• Open Issues: 2
• Metadata Files:
  • Readme: README.md

Lists

• awesome-stars - universome/class-norm - Class Normalization for Continual Zero-Shot Learning (Python)

README

        
# About

This repo contains the code for the [Class Normalization for Continual Zero-Shot Learning paper](https://arxiv.org/abs/2006.11328) from ICLR 2021:

- the code to reproduce ZSL and CZSL results

- the proposed CZSL metrics (located in `src/utils/metrics.py`)

- fast python implementation of the AUSUC metric



[arXiv Paper]

[Google Colab]

[OpenReview Paper]



In this project, we explored different normalization strategies used in ZSL and proposed a new one (class normalization) that is suited for deep attribute embedders.

This allowed us to outperform the existing ZSL model with a simple 3-layer MLP trained just in 30 seconds.

Also, we extended ZSL ideas into a more generalized setting: Continual Zero-Shot Learning, proposed a set of metrics for it and tested several baselines.







# Installation & training

## Data preparation

### For ZSL

For ZSL, we tested our method on the standard GBU datasets which you can download from [the original website](https://www.mpi-inf.mpg.de/departments/computer-vision-and-machine-learning/research/zero-shot-learning/zero-shot-learning-the-good-the-bad-and-the-ugly).

It is the easiest to follow our [Google Colab](class-norm-for-czsl.ipynb) to reproduce the results.

### For CZSL

For CZSL, we tested our method on SUN and CUB datasets.

In contrast to ZSL, in CZSL we used raw images as inputs instead of an ImageNet-pretrained model's features.

For CUB, please follow the instructions in the [A-GEM repo](https://github.com/facebookresearch/agem). Note, that CUB images dataset are now to be downloaded manually from [here](http://www.vision.caltech.edu/visipedia/CUB-200-2011.html), but we used the same splits as A-GEM.

Put the A-GEM splits into the CUB data folder.

For SUN, download the data from the [official website](https://cs.brown.edu/~gmpatter/sunattributes.html), put it under `data/SUN` and then follow the instructions in [scripts/sun_data_preprocessing.py](scripts/sun_data_preprocessing.py)

## Installing the `firelab` dependency

You will need to install [firelab library](https://github.com/universome/firelab) to run the training:

```

pip install firelab

```

## Running ZSL training

Please, refer to this [Google Colab notebook](class-norm-for-czsl.ipynb): it contains the code to reproduce our results.

## Running CZSL training

To run CZSL training you will need to run the command:

```

python src/run.py -c basic|agem|mas|joint -d cub|sun

```

Please note, that by default we load all the data into memory (to speed up things).

This behaviour is controled by the `in_memory` flag in the config.

# Results

## Zero-shot learning results







## Continual Zero-Shot Learning results







## Training speed results for ZSL