3dkeypoints-da's Introduction

Unsupervised Domain Adaptation for 3D Keypoint Estimation via View Consistency

This repository is the PyTorch implementation for the network presented in:

Xingyi Zhou, Arjun Karpur, Chuang Gan, Linjie Luo, Qixing Huang, Unsupervised Domain Adaptation for 3D Keypoint Estimation via View Consistency ECCV 2018(arXiv:1712.05765)

Contact: [email protected]

Requirements

cudnn
PyTorch
Python with h5py, opencv and progress
Optional: tensorboard

Data

The following datasets are used in this repo. If you use the data provided, please also consider citing them:
- ModelNet
- ShapeNet and keypoint annotation provided in SyncSpecCNN.
- Redwood Dataset
Download the pre-processing data and annotations here, and un-zip them on data.

Testing

Download our pre-trained model on Redwood Depth dataset and move it to models.
Run the test.

 python main.py -expID demo -loadModel ../models/Redwood.pth.tar -test

Visualize the results.

python tools/vis.py ../exp/Chair/demo/img_valTarget ../exp/Chair/demo/valTarget.txt

Training

Stage1: Train the source model.

python main.py -expID Source -epochs 120 -dropLR 90

Our results of this stage is provided here.

Stage2: Adapt to the target domain with shape consistency loss.

python main.py -expID Redwood -targetDataset Redwood -targetRatio 1 -shapeWeight 1 -loadModel ../models/ModelNet120.tar -LR 0.01

3dkeypoints-da's People

Contributors

Stargazers

Watchers

3dkeypoints-da's Issues

Requirements.txt

Thank you so much for providing the code.
Although you mention the dependencies in README, the version information, especially for pytorch and torchvision, is missing. It would be very helpful and nice if you could also provide requirements.txt.

Positions of key points under same unit

Hello, thanks for your open code! Are the positions of the key points under the three directions in the same unit? Thank you very much!

``
pts = np.zeros((self.nViews, ref.J, 3), dtype = np.float32)
pts[v] = self.annot[index, vv].copy()
return inp, pts, meta

Backward error in ShapeConsistencyCriterion.py

When I download the datasets and run python main.py -expID Source -epochs 120 -dropLR 90 to train the source model, an error happened.

Environment

Python2.7
PyTorch0.4.0
Torchvision 0.2.2

Traceback

Traceback (most recent call last):
  File "main.py", line 152, in <module>
    main()
  File "main.py", line 98, in main
    epoch)
  File "/home/inoue/experiment/da_regression/3DKeypoints-DA/src/train.py", line 79, in train
    return step(args, 'train', epoch, train_loader, model, optimizer, M=M)
  File "/home/inoue/experiment/da_regression/3DKeypoints-DA/src/train.py", line 65, in step
    loss.backward()
  File "/home/inoue/experiment/.cuda10_pt_0_4_py2_env/local/lib/python2.7/site-packages/torch/tensor.py"
, line 93, in backward
    torch.autograd.backward(self, gradient, retain_graph, create_graph)
  File "/home/inoue/experiment/.cuda10_pt_0_4_py2_env/local/lib/python2.7/site-packages/torch/autograd/_
_init__.py", line 89, in backward
    allow_unreachable=True)  # allow_unreachable flag
  File "/home/inoue/experiment/da_regression/3DKeypoints-DA/src/layers/ShapeConsistencyCriterion.py", li
ne 74, in backward
    g, v]) / ref.J / 3 / self.nViews / G
RuntimeError: expand(torch.FloatTensor{[10, 3]}, size=[30]): the number of sizes provided (1) must be gr
eater or equal to the number of dimensions in the tensor (2)

it seems that there is shape mismatch between grad_input[g * self.nViews + v] and points[g, v] - target[g, v].
Do you have any suggestion to solve the error?

Recommend Projects

xingyizhou / 3dkeypoints-da Goto Github PK

3dkeypoints-da's Introduction

Unsupervised Domain Adaptation for 3D Keypoint Estimation via View Consistency

Requirements

Data

Testing

Training

3dkeypoints-da's People

Contributors

Stargazers

Watchers

Forkers

3dkeypoints-da's Issues

Requirements.txt

Positions of key points under same unit

Backward error in ShapeConsistencyCriterion.py

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent