python-objectness

Python implementation of PAMI 2012 paper "Measuring the Objectness of Image Windows" and CVPR 2010 paper "What is an object ?".

This implementation is based on the original matlab code.

Results

Detailed comparisons to the original Matlab implementation can be found here.

Get started

Clone this repository:

git clone https://github.com/RookieHong/python-objectness.git
cd python-objectness

Install the requirements below.
Run the demo and see the results:
```
python demo.py
```

(Optional, but recommended) A python implementation of superpixels segmentation is used as default, you can do the following to use the official C++ implementation through pybind11 instead (faster and more accurate):

# Compile the python interface
$ sudo apt install cmake   # Install cmake
$ cd c_segment
$ mkdir build
$ cd build
$ cmake .. && make

Then, switch from py_segment to c_segment by modifying these two lines in computeScores.py:

# Comment the py_segment line and uncomment the c_segment line

S = c_segment_img(img=img, sigma=sigma, k=k, min_area=min_area)                         # C segment
# S = py_segment_img(img=img, sigma=sigma, neighbor=4, K=k, min_comp_size=min_area)     # Python segment

Requirements

numpy
opencv-python
easydict
mat4py
numba

Notes

Because the implementation of many functions in MATLAB and numpy is different, the results might be slightly different from the original code.

The superpixel computation code in py-segment folder is modified from a python implementation, so this will also cause differences in code results.

To-do

Simplify the indices to make them more readable.

Citation

If you find this repository useful, consider citing their paper:

@article{alexe2012measuring,
  title={Measuring the objectness of image windows},
  author={Alexe, Bogdan and Deselaers, Thomas and Ferrari, Vittorio},
  journal={IEEE transactions on pattern analysis and machine intelligence},
  volume={34},
  number={11},
  pages={2189--2202},
  year={2012},
  publisher={IEEE}
}

@inproceedings{alexe2010object,
  title={What is an object?},
  author={Alexe, Bogdan and Deselaers, Thomas and Ferrari, Vittorio},
  booktitle={2010 IEEE computer society conference on computer vision and pattern recognition},
  pages={73--80},
  year={2010},
  organization={IEEE}
}