search

ahmetius / LP-DeepSSL

Code for CVPR 2019 paper Label Propagation for Deep Semi-supervised Learning
MIT License
116 stars 17 forks source link

readme

LabelProp-SSDL

This repository contains the code for the following paper:

A. Iscen, G. Tolias, Y. Avrithis, O. Chum. "Label Propagation for Deep Semi-supervised Learning", CVPR 2019

This code generally follows Mean Teacher Pytorch implementation. Python 3 and FAISS are required.

Data pre-processing:

CIFAR-10

As in the original Mean Teacher repository, run the following command:

>> cd data-local/bin
>> ./prepare_cifar10.sh

CIFAR-100

Run the similar command for CIFAR-100, taken from the fastswa-semi-sup repository:

>> cd data-local/bin
>> ./prepare_cifar100.sh

Mini-Imagenet

We took the Mini-Imagenet dataset hosted in this repository and pre-processed it.

Download train.tar.gz and test.tar.gz, and extract them in the following directory:

>> ./data-local/images/miniimagenet/

Running the experiments:

There are two stages to our method.

Stage 1 consists of training a network with known labels only. Following command can be run to reproduce this experiment:

>> python train_stage1.py --exclude-unlabeled=True --num-labeled=$NOLABELS --gpu-id=$GPUID --label-split=$SPLITID --isMT=False --isL2=True --dataset=$DATASET

where $NOLABELS is the number of labeled points, $GPUID is the GPU to be used (0 by default), $SPLITID is the ID of the split to be used (10-19 for cifar10, 10-12 for cifar100 and miniimagenet), $DATASET is the name of the dataset (cifar10, cifar100, or miniimagenet), and $LABELED_IN_BATCH is the number of labeled images in a batch (50 for cifar10, 31 for cifar100 and miniimagenet).

After the training for Stage 1 is completed, run the following command for Stage 2, which resumes the training from the model trained in Stage 1, but this time with pseudo-labels using the entire dataset:

>> python train_stage2.py --labeled-batch-size=$LABELED_IN_BATCH --num-labeled=$NOLABELS --gpu-id=$GPUID --label-split=$SPLITID --isMT=False --isL2=True --dataset=$DATASET

Combining with Mean Teacher:

Use the following commands for combining Mean Teacher with our method:

Stage 1:

>> python train_stage1.py --labeled-batch-size=$LABELED_IN_BATCH --num-labeled=$NOLABELS --gpu-id=$GPUID --label-split=$SPLITID --isMT=True --isL2=False --dataset=$DATASET

Stage 2:

>> python train_stage2.py --labeled-batch-size=$LABELED_IN_BATCH --num-labeled=$NOLABELS --gpu-id=$GPUID --label-split=$SPLITID --isMT=True --isL2=False --dataset=$DATASET