In the field of computer vision, there is often too much redundant information in large images. After preprocessing, the areas of human interest are relatively scattered, which is not conducive to centralized processing. We redeploy these areas to facilitate subsequent centralized processing.

• 1.Requirements

  opencv-python
  numpy
  

• 2.Useage&&Developing

• 1.Building && run

# opencv == 3.3.1
$ mkdir build && cd build
$ cmake .. && make -j8
$ ../bin/roi_layout

• 2.Useage&&Developing

• How to get layout rects? use get_layout().
• How to get layout image? use get_layout_mat().
• How to infer layout result? use TryLayout().
• How to map box in layout image to orignal image? use MapLayoutBoxToROI().

If you want to get a different deployment shape, just fix the following sentence in file.

  /**
   * ROI_layout.cpp
   * void ROILayout::FindBestLocation(Rect &rect)
  **/
  float score = ratio_wh * ratio_area;  // Can be customized.

• C++ version adds gap
• Support merge Adjacent rectangle
• Replace IOU with CIOU
• Expand rectangle

This project is licensed under the MIT License - see the LICENSE.md file for details.

Please cite roi-layout in your publications if it helps your research:

@misc{roi_layout,
    Author = {niovl},
    Title = {{roi-layout: Redeploy the roi Extracted From the Large Image To the Small Image.}},
    Howpublished = {\url{https://github.com/niovl/roi_layout}},
    Year = {2020}
}