search

microsoft / transductive-vos.pytorch

a transductive approach for video object segmentation
155 stars 32 forks source link

Inference without the motion prior #9

Closed juntingzh closed 4 years ago

juntingzh commented 4 years ago

Thanks for sharing such a neat implementation

When I try to understand the effect of the motion prior, I ran inference without the motion prior by commenting off the following lines: https://github.com/microsoft/transductive-vos.pytorch/blob/5146b32c953c3fe90ed3caee67c02ed20dfa8c7e/lib/predict.py#L48-L55

However, the final score was not the same as what was reported in the paper (J=69.0 in Table 2). Here is what I got: +--------------+--------+----------+---------+--------+----------+---------+ | Method | J_mean | J_recall | J_decay | F_mean | F_recall | F_decay | +--------------+--------+----------+---------+--------+----------+---------+ | val-no-motion | 0.618 | 0.714 | 0.224 | 0.675 | 0.772 | 0.253 | +--------------+--------+----------+---------+--------+----------+---------+

Could you please help me understand what went wrong? What was the setting for the ablation study in the paper? Thank you!

eastOffice commented 4 years ago

Hello,

The motion prior corresponds to line 48 - 53, where different sigmas are applied to compute different weights. In your case, no weights are applied and a worse result is expected.