antonilo/unsupervised_detection

image size predicted mask on FBMS

fyangneil opened this issue · 5 comments

The provided FBMS results is not consistent with the ground truth image in terms of image size. For example, the results image of the first frame of cars 10 is 384 192 but the corresponding ground truth image size is 640480.

dear @fyangneil,

there are several re-shaping operations in post-processing to make sure that the image resolution is as expected. I would encourage you to check out the readme for details.

@antonilo do you have the test script for FBMS dataset?

@antonilo I use the provided model and test it on FBMS test dataset, the result is 0.498. The script I used is test_DAVIS2016_raw.sh with corresponding modification for FBMS dataset. I also tried the post processing code but the result is not good, either. I am not sure what the problem is?

please follow the instructions and make sure to reproduce the same results as in here http://rpg.ifi.uzh.ch/data/detection_results.zip