About the evaluation metric mIoU.

[forrestsz]

Hi, thank for your nice work and job. After reading your code, i have a little question. can you share the code how to compute the mIoU in unsupervised way, because which is different from supervised segmentation. Thank you!👍

[kanezaki]

Thanks for your comment! The code is coming soon. The sample code and segments are currently in my Dropbox link here.

$ python miou_calculation.py --bsd500 --mode 1 --gt ./gt --input ./results

mode 1: "all" mode 2: "fine" mode 3: "coarse"

Before mIOU calculation, you need to distinguish unconnected segments that have the same class labels. I used this library to relabel connected components.

[forrestsz]

great thank for your reply!

[euwern]

Thanks for your comment! The code is coming soon. The sample code and segments are currently in my Dropbox link here.

$ python miou_calculation.py --bsd500 --mode 1 --gt ./gt --input ./results

mode 1: "all" mode 2: "fine" mode 3: "coarse"

Before mIOU calculation, you need to distinguish unconnected segments that have the same class labels. I used this library to relabel connected components.

Can you describe how you use the library (mentioned above) to relabel connected components? Assuming that I have just cloned the repo and compile the library, what's the next step? It will be helpful if you can share the command line to relabel the connected components.

[kanezaki]

I used the function SuperpixelTools::relabelConnectedSuperpixels(segmentation) in examples/cpp/evaluation.cpp. To obtain the output, I inserted the following lines in L151 and then used the executable bin/example_evaluation.

    for (int i = 0; i < segmentation.rows; ++i) {
        for (int j = 0; j < segmentation.cols-1; ++j) {
          std::cerr << segmentation.at<int>(i, j) << ",";
        }
        std::cerr << segmentation.at<int>(i, segmentation.cols-1) << std::endl;
    }

[charleneolive]

Regarding your code, what is the use of these lines?

gg = np.zeros(gt_array.shape) gt_mask = np.where(gt_array > 0, 1, gg) determinant_array = gt_mask * input_array label_list = np.unique(gt_array)

I saw that you have defined label_list already above.

[lutfiex]

Thanks for your comment! The code is coming soon. The sample code and segments are currently in my Dropbox link here.

$ python miou_calculation.py --bsd500 --mode 1 --gt ./gt --input ./results

mode 1: "all" mode 2: "fine" mode 3: "coarse"

Before mIOU calculation, you need to distinguish unconnected segments that have the same class labels. I used this library to relabel connected components.

have the other method to relabel connected components?