Estimated Depth Map Helps Image Classification

GitHub - yihui-he/Estimated-Depth-Map-Helps-Image-Classification: Depth estimation with neural network, and learning on RGBD images

Estimated Depth Map Helps Image Classification

if you find our work helpful in your research, please consider citing:

@article{estimated2017he,
  title={Estimated Depth Map Helps Image Classification},
  author={He, Yihui},
  journal={arXiv preprint arXiv:1709.07077},
  year={2017}
}

how to test

1. you can run tryhere.ipynb to test performance on RGBD and RGB images.
2. you can do depth estimation in train/ folder

Approach

1. train a mapping map RGB to depth
2. convert cifar10 to images
3. convert RGBD to cifar10 format
4. train neural network on RGBD dataset

download

dcnf-fcspRGBD CIFAR-10

External Links

all dataset descriptionsa good one