Asako Kanezaki

Hi, Please see #2 for our evaluation code.

Yup, the iou with a blank image will be rather high when the ground truth segments are large. Here are the comparison of mIOU with a blank image and our...

A blank image can be regarded as a single large segment. IoU of a large segment and a small segment is low (near zero).

Thanks for your comment! The code is coming soon. The sample code and segments are currently in my Dropbox link [here](https://www.dropbox.com/s/jhhbb5r174hj2jq/miou_calculation_example.zip?dl=0). ``` $ python miou_calculation.py --bsd500 --mode 1 --gt ./gt...

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;...

The saving function is included in "demo_ref.py" in [our new repository,](https://github.com/kanezaki/pytorch-unsupervised-segmentation-tip) so please refer there. Thanks!

It's done in my Caffe implementation [here](https://github.com/kanezaki/rotationnet#test-pose-estimation), but not yet done in this PyTorch implementation. Please refer there for the process of aligning/matching views.

I uploaded the necessary file here: https://www.dropbox.com/s/9h7xel1nl5gc0lc/ilsvrc_2012_train_iter_310k?dl=0 Thanks!

I'm not sure about that error... Have you compiled [caffe-rotationnet2](https://github.com/kanezaki/caffe-rotationnet2) with no error?

Understood. These lines in _Makefile.config_ should be left commented out. ``` # uncomment to disable IO dependencies and corresponding data layers # USE_OPENCV := 0 # USE_LEVELDB := 0 #...