densemapnet

KO | EN

gitlite — search

#javascript #ai #typescript #golang #go #macos #java #kubernetes #awesome-list #security #mcp #ios #openai #open-source

(★ 65)

Keras code of my ICRA 2018 paper "Fast Disparity Estimation using Dense Networks"

File Explorer

Download Latest Version (.zip)

- driving.densemapnet.weights.h5
- flying.densemapnet.weights.h5
- kitti2015.densemapnet.weights.h5
- monkaa.densemapnet.weights.h5
- mpi.densemapnet.weights.h5
- Driving-Monkaa-Flying.png
- Driving.png
- ICRA2018_Poster_A4.pdf
- KITTI2015.png
.gitignore
densemapnet.py
LICENSE
predictor.py
README.md
utils.py

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# DenseMapNet
Keras code of **_"Fast Disparity Estimation using Dense Networks"_** in the proceedings of [International Conference on Robotics and Automation, Australia, 2018 (ICRA 2018)](http://icra2018.org/).

The paper can be downloaded [here](https://arxiv.org/pdf/1805.07499.pdf).

**_DenseMapNet_ Features**
- Predicts disparity map using full resolution stereo RGB
- Fast at >=30Hz on NVIDIA 1080Ti GPU
- Tiny network with only 290k parameters 
- Accurate with Low End-Point-Error or EPE

### Sample Predictions ###
#### Driving, Monkaa, and Flying Datasets ####
![Driving, Monkaa and Flying datasets](https://raw.githubusercontent.com/roatienza/densemapnet/master/media/Driving-Monkaa-Flying.png)
#### KITTI 2015 ####
![KITTI 2015](https://raw.githubusercontent.com/roatienza/densemapnet/master/media/KITTI2015.png)

## Demo

## Dataset
Download datasets:
1. [`driving`](https://drive.google.com/file/d/1q01ffNwvnZkrdw58_LIX-tf-vkzsGGmI/view?usp=sharing)
2. [`mpi`](https://drive.google.com/file/d/1mntUmDxpmCPafYh9nCDWPgT6JyzVovDK/view?usp=sharing)

Copy: `cp driving.tar.bz2 densemapnet/dataset`

Change dir and extract: `cd densemanpnet/dataset; tar jxvf driving.tar.bz2`

Available datasets:

1. `driving` - [Driving](https://lmb.informatik.uni-freiburg.de/resources/datasets/SceneFlowDatasets.en.html) 
2. `mpi` - [MPI Sintel](http://sintel.is.tue.mpg.de/)

Additional datasets will be available in the future.

## Training
In some datasets, the train data is split into multiple files. For example, `driving` is split into 4 files while `mpi` fits into 1 file.

To train the network:

`python3 predictor.py --dataset=driving --num_dataset=4`

Alterntaively, load the pre-trained weigths:

`python3 predictor.py --dataset=driving --num_dataset=4 --weights=checkpoint/driving.densemapnet.weights.h5`

## Testing

To measure EPE using test set:

`python3 predictor.py --dataset=driving --num_dataset=4 --weights=checkpoint/driving.densemapnet.weights.h5 --notrain`

To benchmark speed only:

`python3 predictor.py --dataset=driving --num_dataset=4 --weights=checkpoint/driving.densemapnet.weights.h5 --predict`

To generate disparity predictions on both train and test datasets (complete sequential images used to create the video):

`python3 predictor.py --dataset=driving --num_dataset=4 --weights=checkpoint/driving.densemapnet.weights.h5 --predict
--images`

## Citation
If you find this work useful, please cite:

```
@inproceedings{atienza2018fast,
  title={Fast Disparity Estimation using Dense Networks},
  author={Atienza, Rowel},
  booktitle={2018 IEEE International Conference on Robotics and Automation (ICRA)},
  pages={3207--3212},
  year={2018},
  organization={IEEE}
}
```

densemapnet

File Explorer

Today's picks

Was this content helpful?

Recent Feedback

densemapnet

File Explorer

Today's picks

Was this content helpful?

Recent Feedback

Download README