search

Jack-zz-ze / HFSNet

0 stars 0 forks source link

Questions about the edge loss function. #1

Open WanSheng-SZU opened 6 months ago

WanSheng-SZU commented 6 months ago

Thanks for your work, may I ask why self.loss_fun2(HSI1, 1 - DSM) in HFSNet.py is subtracting DSM with 1, and it seems that DSM is not a tensor that contains only 0 or 1, but the DiceLoss function is set to mode='binary', so shouldn't DSM contain only 0 or 1? Looking forward to your answer, thanks.

Jack-zz-ze commented 6 months ago

In self.loss_fun2, “DSM” is the output of the Canny operation, and its data format contains only o and 1.

WanSheng-SZU commented 6 months ago

@Jack-zz-ze May I ask why the result is also irreducible even after setting a random seed?

Jack-zz-ze commented 6 months ago

You said the result on the Trento dataset?

---Original--- From: @.> Date: Mon, Mar 18, 2024 16:47 PM To: @.>; Cc: @.**@.>; Subject: Re: [Jack-zz-ze/HFSNet] Questions about the edge loss function. (Issue #1)

@Jack-zz-ze May I ask why the result is also irreducible even after setting a random seed?

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

WanSheng-SZU commented 6 months ago

You said the result on the Trento dataset? … ---Original--- From: @.> Date: Mon, Mar 18, 2024 16:47 PM To: @.>; Cc: @.**@.>; Subject: Re: [Jack-zz-ze/HFSNet] Questions about the edge loss function. (Issue #1) @Jack-zz-ze May I ask why the result is also irreducible even after setting a random seed? — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: @.***>

Yes, the results were found to be non-reproducible.

Jack-zz-ze commented 6 months ago

What's the result you get?

---Original--- From: @.> Date: Mon, Mar 18, 2024 18:04 PM To: @.>; Cc: @.**@.>; Subject: Re: [Jack-zz-ze/HFSNet] Questions about the edge loss function. (Issue #1)

You said the result on the Trento dataset? … ---Original--- From: @.> Date: Mon, Mar 18, 2024 16:47 PM To: @.>; Cc: @.@.>; Subject: Re: [Jack-zz-ze/HFSNet] Questions about the edge loss function. (Issue #1) @Jack-zz-ze May I ask why the result is also irreducible even after setting a random seed? — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: @.***>

Yes, the results were found to be non-reproducible.

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

Jack-zz-ze commented 6 months ago

Due to the randomness of the parameter initialization, data divisions, different equipment and so on, the experimental results might have slightly different. You can try different parameter on your device. I also try more time to get the best result. Thank you for your attention and advice!

---Original--- From: @.> Date: Mon, Mar 18, 2024 18:04 PM To: @.>; Cc: @.**@.>; Subject: Re: [Jack-zz-ze/HFSNet] Questions about the edge loss function. (Issue #1)

You said the result on the Trento dataset? … ---Original--- From: @.> Date: Mon, Mar 18, 2024 16:47 PM To: @.>; Cc: @.@.>; Subject: Re: [Jack-zz-ze/HFSNet] Questions about the edge loss function. (Issue #1) @Jack-zz-ze May I ask why the result is also irreducible even after setting a random seed? — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: @.***>

Yes, the results were found to be non-reproducible.

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