search

YangZhang4065 / AdaptationSeg

Curriculum Domain Adaptation for Semantic Segmentation of Urban Scenes, ICCV 2017
129 stars 28 forks source link

Inconsistent Label Resizing during evaluation #2

Open swamiviv opened 7 years ago

swamiviv commented 7 years ago

Hi, Nice work! I have a query regarding your evaluation.

In your paper you have mentioned that: " Since we have to resize the images before feeding them to the segmentation network, we resize the output segmentation mask back to the original image size before running the evaluation against the groundtruth annotations. "

However, in warp_data.py which is called by your eval. code, it seems like the label is also resized to (320,640).

Can you please clarify this inconsistency for me ? Thanks!

YangZhang4065 commented 7 years ago

Thank you for your interest! As mentioned in the 'note' section at the end of our GitHub, this is actually a Keras 2 reimplementation of original Keras 1 version in our ICCV paper. We also cleaned up the messy code. This only serves as a demo purpose.

In the original evaluation code which we used in the ICCV paper, we have a set of another load/resize code to make sure the prediction masks are resized back to the original size.