search

nikhil0209 / Image-Emotion-Classification

Trained an attention-based CNN model to classify images into pre-defined emotion classes ( Amusement, Anger, Awe, Contentment, Disgust, Excitement, Fear and Sadness) based on what emotions humans are likely to link the image with.
0 stars 0 forks source link

Attention mechanism #1

Open efabb95 opened 4 years ago

efabb95 commented 4 years ago

Hi Nikhil, I'm italian student. I'm attending the university of Milan-Bicocca. Can I ask you an info about the attention mechanism? Thanks

nikhil0209 commented 4 years ago

Hi Elia, yes please tell me what info you need about the attention mechanism. Thanks!

On Sat, 18 Apr 2020 at 8:37 PM, Elia Fabbris notifications@github.com wrote:

Hi Nikhil, I'm italian student. I'm attending the university of Milan-Bicocca. Can I ask you an info about the attention mechanism? Thanks

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/nikhil0209/Image-Emotion-Classification/issues/1, or unsubscribe https://github.com/notifications/unsubscribe-auth/AC3WZJ7FI5PGXDBLB5FBIEDRNG647ANCNFSM4MLMNC2Q .

efabb95 commented 4 years ago

Thanks for reply! I can't understand the right components of this mechanism. The matrix W is the input of your attention mechanism, right? The dimensions are 66256. I suppose that X identifying the vector of 256 features, dimension 1*256. Is that right?

efabb95 commented 4 years ago

I'm trying to understand the components but I can't find the right solution.

On apr 18 2020, at 5:57 pm, nikhil0209 notifications@github.com wrote:

Hi Elia, yes please tell me what info you need about the attention mechanism. Thanks!

On Sat, 18 Apr 2020 at 8:37 PM, Elia Fabbris notifications@github.com wrote:

Hi Nikhil, I'm italian student. I'm attending the university of Milan-Bicocca. Can I ask you an info about the attention mechanism? Thanks

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/nikhil0209/Image-Emotion-Classification/issues/1, or unsubscribe https://github.com/notifications/unsubscribe-auth/AC3WZJ7FI5PGXDBLB5FBIEDRNG647ANCNFSM4MLMNC2Q .

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub (https://github.com/nikhil0209/Image-Emotion-Classification/issues/1#issuecomment-615894272), or unsubscribe (https://github.com/notifications/unsubscribe-auth/AJL6EYUABQU4JUR7OYPUVOTRNHEYNANCNFSM4MLMNC2Q).

nikhil0209 commented 4 years ago

I’ll have to go through the paper once, Elia, before I can say anything. I’ll have a look tonight. Thanks!

On Mon, 20 Apr 2020 at 2:43 PM, Elia Fabbris notifications@github.com wrote:

I'm trying to understand the components but I can't find the right solution.

On apr 18 2020, at 5:57 pm, nikhil0209 notifications@github.com wrote:

Hi Elia, yes please tell me what info you need about the attention mechanism. Thanks!

On Sat, 18 Apr 2020 at 8:37 PM, Elia Fabbris notifications@github.com wrote:

Hi Nikhil, I'm italian student. I'm attending the university of Milan-Bicocca. Can I ask you an info about the attention mechanism? Thanks

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/nikhil0209/Image-Emotion-Classification/issues/1, or unsubscribe < https://github.com/notifications/unsubscribe-auth/AC3WZJ7FI5PGXDBLB5FBIEDRNG647ANCNFSM4MLMNC2Q

.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub ( https://github.com/nikhil0209/Image-Emotion-Classification/issues/1#issuecomment-615894272), or unsubscribe ( https://github.com/notifications/unsubscribe-auth/AJL6EYUABQU4JUR7OYPUVOTRNHEYNANCNFSM4MLMNC2Q ).

— You are receiving this because you commented.

Reply to this email directly, view it on GitHub https://github.com/nikhil0209/Image-Emotion-Classification/issues/1#issuecomment-616418247, or unsubscribe https://github.com/notifications/unsubscribe-auth/AC3WZJZN3NWKMSCMNQ2X2MLRNQG2TANCNFSM4MLMNC2Q .

efabb95 commented 4 years ago

Thanks a lot Nikhil, you are very gentle! If you want, you can contact me on Twitter. See you soon!

nikhil0209 commented 4 years ago

Hi Elia,

I have been really busy with work this week. I'll definitely have a look as soon as I can.

Regards, Nikhil

On Mon, 20 Apr 2020 at 7:21 PM, Elia Fabbris notifications@github.com wrote:

Thanks a lot Nikhil, you are very gentle! If you want, you can contact me on Twitter. See you soon!

— You are receiving this because you commented.

Reply to this email directly, view it on GitHub https://github.com/nikhil0209/Image-Emotion-Classification/issues/1#issuecomment-616567603, or unsubscribe https://github.com/notifications/unsubscribe-auth/AC3WZJYR32INGTZJGJMRZATRNRHOPANCNFSM4MLMNC2Q .

efabb95 commented 4 years ago

Ok, thank you.

efabb95 commented 4 years ago

Hi Nikhil, sorry to bother you again, are there news?

Regards, Elia

efabb95 commented 4 years ago

I understood something but only one question: how do you learn the matrix W? Is it composed by random values?