search

EricFH / SOR

implementation of "Salient Object Ranking with Position-Preserved Attention"
Apache License 2.0
25 stars 5 forks source link

“inference”，“Dataset” #5

Open cherryolg opened 2 years ago

cherryolg commented 2 years ago

Hello, author. I want to apply your model to other test datasets, not ASR. Does your model have any requirements for the test dataset? Do I need to have all the labels of ASR? Looking forward to your reply!

EricFH commented 2 years ago

Inference stage does not need label.

cherryolg commented 2 years ago

Thank you very much for your reply. I want to make sure what you mean: can I use any dataset I want to use for inference? Just need to include "images" in the dataset, right?Looking forward to  your reply again, thanks!

------------------ 原始邮件 ------------------ 发件人: "EricFH/SOR" @.>; 发送时间: 2022年3月25日(星期五) 晚上10:07 @.>; @.**@.>; 主题: Re: [EricFH/SOR] “inference”，“Dataset” (Issue #5)

Inference stage does not need label.

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

cherryolg commented 2 years ago

Hello, author!

For example, my test dataset size is not (480,640). I found that when considering "pos" in the code, i.e., positional coordinate maps, the size was fixed (480,640). And paper says “We take an end-to-end training strategy. The shape of input images is 640 X 480, Which is the same as the original images." Can this model process image whose size is not (480,640)? The actual size may be smaller than (480,640).

------------------ 原始邮件 ------------------ 发件人: "EricFH/SOR" @.>; 发送时间: 2022年3月25日(星期五) 晚上10:07 @.>; @.**@.>; 主题: Re: [EricFH/SOR] “inference”，“Dataset” (Issue #5)

Inference stage does not need label.

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

EricFH commented 2 years ago

The size is not fixed, every proposal will get the same size after roi pooling no matter what input shape it is. inference stage does not need label. Labels in validate set is for evaluation.

cherryolg commented 2 years ago

Thank you very much! Your answer helped me a lot!

------------------ 原始邮件 ------------------ 发件人: "EricFH/SOR" @.>; 发送时间: 2022年3月29日(星期二) 中午1:39 @.>; @.**@.>; 主题: Re: [EricFH/SOR] “inference”，“Dataset” (Issue #5)

The size is not fixed, every proposal will get the same size after roi pooling no matter what input shape it is. inference stage does not need label. Labels in validate set is for evaluation.

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

jiangyuanlin01 commented 1 year ago

Can you provide the code of the final results with different colors？