Skip to content
Learning to Separate Object Sounds by Watching Unlabeled Video (ECCV 2018)
Python
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
data
models
options
util
README.md
test.py
train.py

README.md

Learning to Separate Object Sounds by Watching Unlabeled Video

Learning to Separate Object Sounds by Watching Unlabeled Video: [Project Page] [arXiv]

This repository contains the deep MIML network implementation for our ECCV 2018 paper.

If you find our code or project useful in your research, please cite:

    @inproceedings{gao2018objectSounds,
      title={Learning to Separate Object Sounds by Watching Unlabeled Video},
      author={Gao, Ruohan and Feris, Rogerio and Grauman, Kristen},
      booktitle={ECCV},
      year={2018}
    }

Use the following command to train the deep MIML network:

python train.py --HDF5FileRoot /your_hdf5_file_root --name deepMIML --checkpoints_dir checkpoints --model MIML --batchSize 256 --learning_rate 0.001 --learning_rate_decrease_itr 5 --decay_factor 0.94 --display_freq 10 --save_epoch_freq 5 --save_latest_freq 500 --gpu_ids 0 --nThreads 2 --num_of_fc 1 --with_batchnorm --continue_train --niter 300 --L 15 --validation_on --validation_freq 50 --validation_batches 10 --selected_classes --using_multi_labels |& tee -a train.log

Acknowlegements

Our code borrows heavily from the the CycleGAN implementation https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix/.

You can’t perform that action at this time.