Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/MightyChaos/LKVOLearner
Learning Depth from Monocular Videos using Direct Methods, CVPR 2018
https://github.com/MightyChaos/LKVOLearner
ddvo depth-estimation posenet-ddvo pytorch unsupervised-learning visual-odometry
Last synced: 6 days ago
JSON representation
Learning Depth from Monocular Videos using Direct Methods, CVPR 2018
- Host: GitHub
- URL: https://github.com/MightyChaos/LKVOLearner
- Owner: MightyChaos
- License: bsd-3-clause
- Created: 2018-01-03T01:59:18.000Z (almost 7 years ago)
- Default Branch: master
- Last Pushed: 2018-11-22T07:53:07.000Z (almost 6 years ago)
- Last Synced: 2024-08-02T08:10:07.522Z (3 months ago)
- Topics: ddvo, depth-estimation, posenet-ddvo, pytorch, unsupervised-learning, visual-odometry
- Language: Python
- Homepage:
- Size: 74.2 KB
- Stars: 230
- Watchers: 13
- Forks: 38
- Open Issues: 6
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
- awesome-depth-estimation - Code
- awesome-depth - 4
README
# Learning Depth from Monocular Videos using Direct Methods
Implementation of the methods in "[Learning Depth from Monocular Videos using Direct Methods](http://openaccess.thecvf.com/content_cvpr_2018/papers/Wang_Learning_Depth_From_CVPR_2018_paper.pdf)".
If you find this code useful, please cite our paper:
```
@InProceedings{Wang_2018_CVPR,
author = {Wang, Chaoyang and Miguel Buenaposada, José and Zhu, Rui and Lucey, Simon},
title = {Learning Depth From Monocular Videos Using Direct Methods},
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2018}
}
```
## Dependencies
- Python 3.6
- PyTorch 0.3.1 (latter or eariler version of Pytorch is non-compatible.)
- visdom, dominate
## Training
### data preparation
We refer "[SfMLeaner](https://github.com/tinghuiz/SfMLearner)" to prepare the training data from KITTI. We assume the processed data is put in directory "./data_kitti/".
### training with different pose prediction modules
Start visdom server before for inspecting learning progress before starting the training process.
```
python -m visdom.server -port 8009
```
1. #### train from scratch with PoseNet
```
bash run_train_posenet.sh
```
see [run_train_posenet.sh](https://github.com/MightyChaos/LKVOLearner/blob/master/run_train_posenet.sh) for details.
2. #### finetune with DDVO
Use pretrained posenet to give initialization for DDVO. Corresponds to the results reported as "PoseNet+DDVO" in the paper.
```
bash run_train_finetune.sh
```
see [run_train_finetune.sh](https://github.com/MightyChaos/LKVOLearner/blob/master/run_train_finetune.sh) for details.
## Testing
- Pretrained depth network reported as "Posenet-DDVO(CS+K)" in the paper [[download](https://drive.google.com/file/d/1SJWLfA7kqpERj_U2gYXl7Vuy1eQyOO_K/view?usp=sharing)].
- Depth prediction results on KITTI eigen test split(see Table 1 in the paper): [[Posenet(K)](https://drive.google.com/open?id=1Wj7ulSimrvrzNx4TRd-JspmX3DJwgPiV)], [[DDVO(K)](https://drive.google.com/open?id=1wiODwgX_Vm_w7fVK1y_X5CNJTtgaPwcN)], [[Posenet+DDVO(K)](https://drive.google.com/open?id=1uUQJLcUOoY2hG6QS_F-wbM3GDAjD-Z5h)],[[Posenet+DDVO(CS+K)](https://drive.google.com/open?id=1hp4zFgK5NSNGdvaQL2ZumeinMQY_-AwK)]
- To test yourself:
```
CUDA_VISIBLE_DEVICES=0 nice -10 python src/testKITTI.py --dataset_root $DATAROOT --ckpt_file $CKPT --output_path $OUTPUT --test_file_list test_files_eigen.txt
```
## Evaluation
We again refer to "[SfMLeaner](https://github.com/tinghuiz/SfMLearner)" for their evaluation code.
## Acknowledgement
Part of the code structure is borrowed from "[Pytorch CycleGAN](https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix)"