search

Xingxu1996 / AIR_iccv2019

coder for International Conference on Computer Vision 2019 paper "APSE for Affective Image Retrieval"
8 stars 3 forks source link

Npair Loss coding?? #2

Open VIROBO-15 opened 4 years ago

VIROBO-15 commented 4 years ago

is the Npair loss coding correct as i am getting many errors in it. can u please show it to me is it running or not.

Xingxu1996 commented 4 years ago

Please show the errors. The code is running well.

VIROBO-15 commented 4 years ago

in line numbet 53. can u tell me the dimensions of n2 and n4 ???

On Wed, Sep 23, 2020, 3:37 PM Xingxu1996 notifications@github.com wrote:

Please show the errors. The code is running well.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/Xingxu1996/AIR_iccv2019/issues/2#issuecomment-697267064, or unsubscribe https://github.com/notifications/unsubscribe-auth/ANQL5YEVX7YETZEVORTYCMLSHHCHXANCNFSM4RWVDPWQ .

Xingxu1996 commented 4 years ago

2 and 8

yxx

yxx_hbgd@163.com | 签名由网易邮箱大师定制 On 9/23/2020 19:22，Amandeep Kumarnotifications@github.com wrote：

in line numbet 53. can u tell me the dimensions of n2 and n4 ???

On Wed, Sep 23, 2020, 3:37 PM Xingxu1996 notifications@github.com wrote:

Please show the errors. The code is running well.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/Xingxu1996/AIR_iccv2019/issues/2#issuecomment-697267064, or unsubscribe https://github.com/notifications/unsubscribe-auth/ANQL5YEVX7YETZEVORTYCMLSHHCHXANCNFSM4RWVDPWQ .

— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.

VIROBO-15 commented 4 years ago

if the n2 dimension is [bath_size,2] then in new_np1.py file in line 37 it's size is [8,2] then how can the size of the weight_polarity_anchor in line 53 be [8,8]

On Thu, Sep 24, 2020 at 6:53 AM Xingxu1996 notifications@github.com wrote:

2 and 8

yxx

yxx_hbgd@163.com | 签名由网易邮箱大师定制 On 9/23/2020 19:22，Amandeep Kumarnotifications@github.com wrote：

in line numbet 53. can u tell me the dimensions of n2 and n4 ???

On Wed, Sep 23, 2020, 3:37 PM Xingxu1996 notifications@github.com wrote:

Please show the errors. The code is running well.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub < https://github.com/Xingxu1996/AIR_iccv2019/issues/2#issuecomment-697267064 , or unsubscribe < https://github.com/notifications/unsubscribe-auth/ANQL5YEVX7YETZEVORTYCMLSHHCHXANCNFSM4RWVDPWQ

.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/Xingxu1996/AIR_iccv2019/issues/2#issuecomment-698056440, or unsubscribe https://github.com/notifications/unsubscribe-auth/ANQL5YEXH62AZ2ATXB3GDGDSHKNSHANCNFSM4RWVDPWQ .

Xingxu1996 commented 4 years ago

It has been initialized by [8,8], which is the structure of mini-batch. Each position will be replaced by the corresponding weight. If you have any other questions. I am on holiday. Please add me WeChat:ixx1799, if you have any other questions. Thank you!

VIROBO-15 commented 4 years ago

is there any other way to communicate with you?

On Thu, Sep 24, 2020, 7:53 AM Xingxu1996 notifications@github.com wrote:

It has been initialized by [8,8], which is the structure of mini-batch. Each position will be replaced by the corresponding weight. If you have any other questions. I am on holiday. Please add me WeChat:ixx1799, if you have any other questions. Thank you!

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/Xingxu1996/AIR_iccv2019/issues/2#issuecomment-698072302, or unsubscribe https://github.com/notifications/unsubscribe-auth/ANQL5YGH36ZOVOCUKIXZKT3SHKUSTANCNFSM4RWVDPWQ .

VIROBO-15 commented 4 years ago

,but the size of anchor1_score_p is [8,2] line 37. then how can it initialize the weights whose size is [8,8]

On Thu, Sep 24, 2020 at 7:53 AM Xingxu1996 notifications@github.com wrote:

It has been initialized by [8,8], which is the structure of mini-batch. Each position will be replaced by the corresponding weight. If you have any other questions. I am on holiday. Please add me WeChat:ixx1799, if you have any other questions. Thank you!

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/Xingxu1996/AIR_iccv2019/issues/2#issuecomment-698072302, or unsubscribe https://github.com/notifications/unsubscribe-auth/ANQL5YGH36ZOVOCUKIXZKT3SHKUSTANCNFSM4RWVDPWQ .

VIROBO-15 commented 4 years ago

is there any other mode ok communication with you??

On Thu, Sep 24, 2020, 8:25 AM Amandeep Kumar kraju39988@gmail.com wrote:

,but the size of anchor1_score_p is [8,2] line 37. then how can it initialize the weights whose size is [8,8]

On Thu, Sep 24, 2020 at 7:53 AM Xingxu1996 notifications@github.com wrote:

It has been initialized by [8,8], which is the structure of mini-batch. Each position will be replaced by the corresponding weight. If you have any other questions. I am on holiday. Please add me WeChat:ixx1799, if you have any other questions. Thank you!

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/Xingxu1996/AIR_iccv2019/issues/2#issuecomment-698072302, or unsubscribe https://github.com/notifications/unsubscribe-auth/ANQL5YGH36ZOVOCUKIXZKT3SHKUSTANCNFSM4RWVDPWQ .

Xingxu1996 commented 4 years ago

Email: xxhbgd@163.com

yxx

yxx_hbgd@163.com | 签名由网易邮箱大师定制 On 9/24/2020 14:54，Amandeep Kumarnotifications@github.com wrote：

is there any other mode ok communication with you??

On Thu, Sep 24, 2020, 8:25 AM Amandeep Kumar kraju39988@gmail.com wrote:

,but the size of anchor1_score_p is [8,2] line 37. then how can it initialize the weights whose size is [8,8]

On Thu, Sep 24, 2020 at 7:53 AM Xingxu1996 notifications@github.com wrote:

It has been initialized by [8,8], which is the structure of mini-batch. Each position will be replaced by the corresponding weight. If you have any other questions. I am on holiday. Please add me WeChat:ixx1799, if you have any other questions. Thank you!

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/Xingxu1996/AIR_iccv2019/issues/2#issuecomment-698072302, or unsubscribe https://github.com/notifications/unsubscribe-auth/ANQL5YGH36ZOVOCUKIXZKT3SHKUSTANCNFSM4RWVDPWQ .

— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.

VIROBO-15 commented 4 years ago

sir, i am still facing problem in the loss function with its dimension

Xingxu1996 commented 4 years ago

8 represents 8 samples from 8 classes， for an anchor, we hope that compute the polarity score for all the samples of all the classes

sir, i am still facing problem in the loss function with its dimension

VIROBO-15 commented 4 years ago

but the polarity is 0 or 1 and emotion is having 8 classes

On Wed, Oct 7, 2020, 11:20 AM Xingxu1996 notifications@github.com wrote:

8 represents 8 samples from 8 classes， for an anchor, we hope that compute the polarity score for all the samples of all the classes

sir, i am still facing problem in the loss function with its dimension

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/Xingxu1996/AIR_iccv2019/issues/2#issuecomment-704708974, or unsubscribe https://github.com/notifications/unsubscribe-auth/ANQL5YEG36XM7YIL4CEDSW3SJP6THANCNFSM4RWVDPWQ .

Xingxu1996 commented 4 years ago

but the polarity is 0 or 1 and emotion is having 8 classes … On Wed, Oct 7, 2020, 11:20 AM Xingxu1996 @.***> wrote: 8 represents 8 samples from 8 classes， for an anchor, we hope that compute the polarity score for all the samples of all the classes sir, i am still facing problem in the loss function with its dimension — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#2 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/ANQL5YEG36XM7YIL4CEDSW3SJP6THANCNFSM4RWVDPWQ .

8 represents 8 samples not 8 classes, "weight_polarity_anchor1[i] = anchor1_score_p[i, polarity.squeeze().data.cpu().numpy()]" selects the polarity score on polarity-wise target

VIROBO-15 commented 4 years ago

anchor1_score_p size is (8,2) and polarity size is (8) Error is mismatching of the matrices

On Wed, Oct 7, 2020, 12:06 PM Xingxu1996 notifications@github.com wrote:

but the polarity is 0 or 1 and emotion is having 8 classes … <#m5055095085368421642> On Wed, Oct 7, 2020, 11:20 AM Xingxu1996 @.***> wrote: 8 represents 8 samples from 8 classes， for an anchor, we hope that compute the polarity score for all the samples of all the classes sir, i am still facing problem in the loss function with its dimension — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#2 (comment) https://github.com/Xingxu1996/AIR_iccv2019/issues/2#issuecomment-704708974>, or unsubscribe https://github.com/notifications/unsubscribe-auth/ANQL5YEG36XM7YIL4CEDSW3SJP6THANCNFSM4RWVDPWQ .

8 represents 8 samples not 8 classes, "weight_polarity_anchor1[i] = anchor1_score_p[i, polarity.squeeze().data.cpu().numpy()]" selects the polarity score on polarity-wise target

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/Xingxu1996/AIR_iccv2019/issues/2#issuecomment-704727635, or unsubscribe https://github.com/notifications/unsubscribe-auth/ANQL5YH4PYPL6FAH27E24IDSJQD7BANCNFSM4RWVDPWQ .