MengyangPu / EDTER

EDTER: Edge Detection with Transformer, in CVPR 2022
MIT License
268 stars 32 forks source link

An error raised during loading NYUD pretrained model #41

Closed Li-yachuan closed 1 year ago

Li-yachuan commented 1 year ago

Thanks for your amazing work at first. But I meet an error when test the NYUD pretrained model. I hope I could get your help. I have tested the BSDS pretrained model successfully. But when test NYUD pretrained model you provide, an error raises. I am not proficient in mmseg, so I cannot solve it independently. I am looking forward to your reply. Thank you very much image

eleveneee commented 1 year ago

Bro, have you solved this problem? This problem occurred when I used the BSDS pre-training model. I am looking forward to your reply. Thank you very much!

Li-yachuan commented 1 year ago

We can use bilinear to make the smaller one of all the matrices whose size is not the same as the larger one.

---Original-- From: @.> Date: Thu, Dec 29, 2022 10:16 AM To: @.>; Cc: @.**@.>; Subject: Re: [MengyangPu/EDTER] An error raised during loading NYUD pretrainedmodel (Issue #41)

Bro, have you solved this problem? This problem occurred when I used the BSDS pre-training model. I am looking forward to your reply. Thank you very much!

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

eleveneee commented 1 year ago

Thank you very much for your reply! Can you be more specific? Due to limited knowledge, I am not able to understand, and look forward to your reply!

程伊 @.***

 

------------------ 原始邮件 ------------------ 发件人: "MengyangPu/EDTER" @.>; 发送时间: 2022年12月29日(星期四) 中午11:48 @.>; @.**@.>; 主题: Re: [MengyangPu/EDTER] An error raised during loading NYUD pretrained model (Issue #41)

We can use bilinear to make the smaller one of all the matrices whose size is not the same as the larger one.

---Original-- From: @.> Date: Thu, Dec 29, 2022 10:16 AM To: @.>; Cc: @.**@.>; Subject: Re: [MengyangPu/EDTER] An error raised during loading NYUD pretrainedmodel (Issue #41)

Bro, have you solved this problem? This problem occurred when I used the BSDS pre-training model. I am looking forward to your reply. Thank you very much!

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

Li-yachuan commented 1 year ago

mmseg/models/segmentors/encoder_decoder.py 388行左右 preds=img.new_zeros((batch_size,64,h_img,w_img)) 适用于nyud

preds=img.new_zeros((batch_size,128,h_img,w_img))

适用于bsds

---Original--- From: @.> Date: Thu, Dec 29, 2022 13:24 PM To: @.>; Cc: @.**@.>; Subject: Re: [MengyangPu/EDTER] An error raised during loading NYUD pretrained model (Issue #41)

Thank you very much for your reply! Can you be more specific? Due to limited knowledge, I am not able to understand, and look forward to your reply!

程伊 @.***

 

------------------ 原始邮件 ------------------ 发件人: "MengyangPu/EDTER" @.>; 发送时间: 2022年12月29日(星期四) 中午11:48 @.>; @.**@.>; 主题: Re: [MengyangPu/EDTER] An error raised during loading NYUD pretrained model (Issue #41)

We can use bilinear to make the smaller one of all the matrices whose size is not the same as the larger one.

---Original--
From: @.>
Date: Thu, Dec 29, 2022 10:16 AM
To:
@.>;
Cc: @.**@.>;
Subject: Re: [MengyangPu/EDTER] An error raised during loading NYUD pretrainedmodel (Issue #41)

Bro, have you solved this problem? This problem occurred when I used the BSDS pre-training model. I am looking forward to your reply. Thank you very much!


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

eleveneee commented 1 year ago

谢谢您的回复!!       我再尝试一下!

程伊 @.***

 

------------------ 原始邮件 ------------------ 发件人: "MengyangPu/EDTER" @.>; 发送时间: 2022年12月29日(星期四) 晚上10:48 @.>; @.**@.>; 主题: Re: [MengyangPu/EDTER] An error raised during loading NYUD pretrained model (Issue #41)

mmseg/models/segmentors/encoder_decoder.py 388行左右 preds=img.new_zeros((batch_size,64,h_img,w_img)) 适用于nyud

preds=img.new_zeros((batch_size,128,h_img,w_img))

适用于bsds

---Original--- From: @.> Date: Thu, Dec 29, 2022 13:24 PM To: @.>; Cc: @.**@.>; Subject: Re: [MengyangPu/EDTER] An error raised during loading NYUD pretrained model (Issue #41)

Thank you very much for your reply! Can you be more specific? Due to limited knowledge, I am not able to understand, and look forward to your reply!

程伊
@.***

 

------------------ 原始邮件 ------------------
发件人: "MengyangPu/EDTER" @.>;
发送时间: 2022年12月29日(星期四) 中午11:48
@.>;
@.**@.>;
主题: Re: [MengyangPu/EDTER] An error raised during loading NYUD pretrained model (Issue #41)

We can use bilinear to make the smaller one of all the matrices whose size is not the same as the larger one.

---Original--
From: @.>
Date: Thu, Dec 29, 2022 10:16 AM
To:
@.>;
Cc: @.**@.>;
Subject: Re: [MengyangPu/EDTER] An error raised during loading NYUD pretrainedmodel (Issue #41)

Bro, have you solved this problem? This problem occurred when I used the BSDS pre-training model. I am looking forward to your reply. Thank you very much!


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

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

Refrain1998 commented 1 year ago

I had the same problem, the command I used was:

python tools/test_local.py --tmpdir nyud2 --globalconfig configs/nyud/EDTER_BIMLA_320x320_40k_nyud_rgb_bs_4.py --global-checkpoint checkpoints/EDTER-NYUD-RGB-StageI.pth --config configs/nyud/EDTER_BIMLA_320x320_40k_nyud_rgb_local8x8_bs_4.py --checkpoint checkpoints/EDTER-NYUD-RGB-StageII.pth

I think "EDTER_BIMLA_320x320_40k_nyud_rgb_local8x8_bs_4.py" does not match "EDTER-NYUD-RGB-StageII.pth".