Middlebury MiddEval3使用教程

只讲最核心：如何把自己的结果上传测评

• README.txt

README.txt

8. Submit your results:

   Our online web interface supports uploading of zip files of results
   on either just the 15 training images or all 30 training and test
   images. You can upload the results on just the training set in
   order to see a temporary table comparing your results with those
   uploaded by others. To do this, run

./makezip Q training Ours

and upload the resulting file resultsQ-Ours.zip using the web
interface. This scripts ensures that all necessary files are
present and that the runtimes files contain numbers.

Once you are ready to submit your results to the permanent table,
you must upload a complete set of results on ALL 30 datasets (15
training, 15 test). All images must have the same resolution (one
of Q, H, or F). You must use constant parameter settings (except
for disaprity ranges), and you cannot “mix and match” resolutions
(e.g., use trainingQ/Piano and trainingH/Teddy). All of this is
ensured if you use “runalg all”, e.g.:

./runalg Q all Ours

Then, create a zip file with all results using

./makezip Q all Ours

and upload the resulting file using the web interface. You will
then be able to provide information about your method and request
publication of both training and test results. Anonynous listings
for double-blind review processes are possible. Note that in order
to prevent parameter tuning to the test data, you will have only
one shot at evaluating your results on the test data. In other
words, you may not request publication for two different sets of
results by the same method. Also, your results on the training data
will be available for download by others. So please be sure to
only submit your final set of results for publication. You may of
course upload and evaluate results on the training set as often as
you want.

按照example：把15幅图像都进行处理，放进相应的位置，makezip指令，然后上传结果测评。

本文来自互联网用户投稿，文章观点仅代表作者本人，不代表本站立场，不承担相关法律责任。如若转载，请注明出处。 如若内容造成侵权/违法违规/事实不符，请点击【内容举报】进行投诉反馈！