Giter VIP home page Giter VIP logo

ice-wacv2018's Introduction

Multi-Task Spatiotemporal Neural Networks for Structured Surface Reconstruction

Introduction

This is a PyTorch implementation for our WACV 2018 paper "Multi-Task Spatiotemporal Neural Networks for Structured Surface Reconstruction".

Alt Text

Note: The pretrained models are trained on the split1 of following larger dataset.

Environment

  • The code is developed with CUDA 9.0, Python >= 3.6, PyTorch >= 1.0

Data Preparation

  1. Download the raw data at ftp://data.cresis.ku.edu/data/rds/2014_Greenland_P3/CSARP_music3D/

    • If you don't want to preprocess the raw data by yourself, please use create_slices.m to generate radar images and convert_mat_to_npy.py to convert them from MATLAB to NumPy files.
  2. If you want to use our dataloaders, please make sure to put the files as the following structure:

    $YOUR_PATH_TO_CRESIS_DATASET
    ├── slices_mat_64x64/
    |   ├── 20140325_05/
    |   |   ├── 001/
    |   |   |   ├── 00001.mat
    |   |   |   ├── ...
    |   |   ├── ...
    │   ├── ...
    |
    ├── slices_npy_64x64/
    |   ├── 20140325_05/
    |   |   ├── 001/
    |   |   |   ├── 00001.npy
    |   |   |   ├── ...
    |   |   ├── ...
    |   ├── ...
    
  3. Create softlinks of datasets:

    cd ice-WACV2018
    ln -s $YOUR_PATH_TO_CRESIS_DATASET data/CReSIS
    ln -s data/target data/CReSIS/target
    

Pretrained Models

  • Download the pretrained models at model_zoo.

Training

  • C3D
cd ice-WACV2018
# Default Hyperparameters
python tools/c3d/train.py
# OR
python tools/c3d/train.py --gpu $CUDA_VISIBLE_DEVICES --batch_size $BS --lr $LR
  • Extract C3D Features
cd ice-WACV2018
# Default Hyperparameters
python tools/c3d/extract_features.py
# OR
python tools/c3d/extract_features.py --gpu $CUDA_VISIBLE_DEVICES --batch_size $BS --checkpoint $C3D_CHECKPOINT
  • RNN
cd ice-WACV2018
# Default Hyperparameters
python tools/rnn/train.py
# OR
python tools/rnn/train.py --gpu $CUDA_VISIBLE_DEVICES --batch_size $BS --lr $LR

Evaluation

cd ice-WACV2018
# Default Hyperparameters
python demo/e2e_eval.py
# OR
python demo/e2e_eval.py --gpu $CUDA_VISIBLE_DEVICES --batch_size $BS --c3d_pth $C3D_CHECKPOINT --rnn_pth $RNN_CHECKPOINT

Citations

If you are using the data/code/model provided here in a publication, please cite our papers:

@inproceedings{icesurface2018wacv,
    title = {Multi-Task Spatiotemporal Neural Networks for Structured Surface Reconstruction},
    author = {Mingze Xu and Chenyou Fan and John D. Paden and Geoffrey C. Fox and David J. Crandall},
    booktitle = {IEEE Winter Conference on Applications of Computer Vision (WACV)},
    year = {2018}
}

@inproceedings{icesurface2017icip, 
    title = {Automatic estimation of ice bottom surfaces from radar imagery},
    author = {Mingze Xu and David J. Crandall and Geoffrey C. Fox and John D. Paden},
    booktitle = {IEEE International Conference on Image Processing (ICIP)},
    year = {2017}
}

ice-wacv2018's People

Contributors

xumingze0308 avatar

Stargazers

davci avatar  avatar Wei Ji avatar Grace Li avatar  avatar  avatar Victor Berger da Silva avatar Mos Ming avatar

Watchers

James Cloos avatar  avatar paper2code - bot avatar

ice-wacv2018's Issues

Some questions about your paper and code

I ran into some problems in reproducing your source code. Should this code be the result of modification after the publication of your paper? There are some differences between your code and your paper. For example, the learnable parameters used in predicting the final result are not reflected in your code. In the comparison of the experimental results in the article, are you using the same 7 frames as before, or the 8 frames you mentioned in the article? May I ask what specific 8 frames of data? Is it the following 8 frames: 20140104_033, 20140104_037, 20140104_039, 20140104_043, 20140104_044, 20140104_045, 20140104_046, 20140104_047? I just use it for academic research experiments, and thank you very much for your help @xumingze0308

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.