SYNDOF (Synthetic Defocus Blur Image Dataset)
_{Paper | Supp | DMENet Repo}

This repository contains the official matlab implementation of SYNDOF generation used in the following paper:

Deep Defocus Map Estimation using Domain Adaptation
Junyong Lee¹, Sungkil Lee², Sunghyun Cho³, and Seungyong Lee^1
1POSTECH, ²Sungkyunkwan University, ³DGIST
IEEE Computer Vision and Pattern Recognition (CVPR) 2019

Picture: Outputs generated from our code– from left to right, synthetic input, defocus map output and defocused image.

Getting Started

Prerequisites

Tested environment

• Download and unzip the synthetic datasets (OneDrive | Dropbox) under ./data:

  ├── data
  │   ├── synthetic_datasets
  │   │   ├── ...
  

SYNDOF Generation

• On matlab console, type

  # max_coc, input_offset, output_offset, is_random_gen, is_gpu, gpu_num
  generate_blur_by_depth(29, 'data', 'out', false, true, 1)

• check the results under ./out, which is structured as,

  ├── ...
  ├── out
  │  ├── blur_map/                    # directory for output defocus map
  │  ├── blur_map_binary/             # directory for binarized defocus map
  │  ├── blur_map_norm/               # directory for normalized defocus map
  │  ├── depth_decomposed/            # directory for decomposed depth
  │  ├── image/                       # directory for input image (with its modified name)
  

Reading SYNDOF

• We rounded real values of defocus map into the nearest 10-th. When you read a defocus map, for example in python, read the file as follows:

  image = (np.float32(cv2.imread(file_name, cv2.IMREAD_UNCHANGED))/10.)[:, :, 1]
  image = image / 7. # 7 = (maxCoC - 1) / 4, where maxCoC is 29 in this case.

Contact

Open an issue for any inquiries. You may also have contact with [email protected]

License

This software is being made available under the terms in the LICENSE file. Any exemptions to these terms require a license from the Pohang University of Science and Technology.

Citation

If you find this code useful, please consider citing:

@InProceedings{Lee2019DMENet,
    author    = {Junyong Lee and Sungkil Lee and Sunghyun Cho and Seungyong Lee},
    title     = {Deep Defocus Map Estimation Using Domain Adaptation},
    booktitle = {IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
    year      = {2019}
}