is episode training used while training

[QuestionLock]

dear jshtok: these days i am working on reproducing the detection results. i am puzzled that i do find episodes generating is done while testing, but i can't found it while end to end training. could you please help me to pick out the episodes generating codes while training, or there is just no episode training. if the later , could you please mind me the training policy. best thanks!

[jshtok]

Hi, we don't use episodes when training, the set of training classes is full and fixed.

On Wed, 26 Aug 2020, 17:27 QuestionLock, notifications@github.com wrote:

dear jshtok: these days i am working on reproducing the detection results. i am puzzled that i do find episodes generating is done while testing, but i can't found it while end to end training. could you please help me to pick out the episodes generating codes while training, or there is just no episode training. if the later , could you please mind me the training policy. best thanks!

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/jshtok/RepMet/issues/32, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACOBU6RRC77T2DJI36OTJ73SCULVDANCNFSM4QL4QK5Q .

[QuestionLock]

Hi, we don't use episodes when training, the set of training classes is full and fixed. … On Wed, 26 Aug 2020, 17:27 QuestionLock, @.***> wrote: dear jshtok: these days i am working on reproducing the detection results. i am puzzled that i do find episodes generating is done while testing, but i can't found it while end to end training. could you please help me to pick out the episodes generating codes while training, or there is just no episode training. if the later , could you please mind me the training policy. best thanks! — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#32>, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACOBU6RRC77T2DJI36OTJ73SCULVDANCNFSM4QL4QK5Q .

thank you for fast replying! i wonder if 100 classes of reps was produced while training so while training the support sets are unnecessary? and how the testing reps of the testing classes are generated using episode testing?

[jshtok]

Correct, there is no support sets, just ordinary batches with GT labels. At test time the reps are computed as kmeans on all the predicted boxes overlapping GT boxes on support samples IoU>.0.7

On Wed, 26 Aug 2020, 17:50 QuestionLock, notifications@github.com wrote:

Hi, we don't use episodes when training, the set of training classes is full and fixed. … <#m7109706435892433455> On Wed, 26 Aug 2020, 17:27 QuestionLock, @.***> wrote: dear jshtok: these days i am working on reproducing the detection results. i am puzzled that i do find episodes generating is done while testing, but i can't found it while end to end training. could you please help me to pick out the episodes generating codes while training, or there is just no episode training. if the later , could you please mind me the training policy. best thanks! — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#32 https://github.com/jshtok/RepMet/issues/32>, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACOBU6RRC77T2DJI36OTJ73SCULVDANCNFSM4QL4QK5Q .

thank you for fast replying! i wonder if 100 classes of reps was produced while training so while training the support sets are unnecessary? and how the testing reps are generated?

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/jshtok/RepMet/issues/32#issuecomment-680928642, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACOBU6X6QF4UQLFYWRELHJTSCUOKTANCNFSM4QL4QK5Q .

[QuestionLock]

great thanks. i get it

------------------ Original ------------------ From: Joseph Shtok <notifications@github.com> Date: Wed,Aug 26,2020 10:53 PM To: jshtok/RepMet <RepMet@noreply.github.com> Cc: QuestionLock <1072031361@qq.com>, Author <author@noreply.github.com> Subject: Re: [jshtok/RepMet] is episode training used while training (#32)

Correct, there is no support sets, just ordinary batches with GT labels. At test time the reps are computed as kmeans on all the predicted boxes overlapping GT boxes on support samples IoU>.0.7

On Wed, 26 Aug 2020, 17:50 QuestionLock, <notifications@github.com> wrote:

> Hi, we don't use episodes when training, the set of training classes is > full and fixed. > … <#m7109706435892433455> > On Wed, 26 Aug 2020, 17:27 QuestionLock, @.***> wrote: dear jshtok: > these days i am working on reproducing the detection results. i am puzzled > that i do find episodes generating is done while testing, but i can't found > it while end to end training. could you please help me to pick out the > episodes generating codes while training, or there is just no episode > training. if the later , could you please mind me the training policy. best > thanks! — You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub <#32 > <https://github.com/jshtok/RepMet/issues/32&gt;&gt;, or unsubscribe > https://github.com/notifications/unsubscribe-auth/ACOBU6RRC77T2DJI36OTJ73SCULVDANCNFSM4QL4QK5Q > . > > thank you for fast replying! i wonder if 100 classes of reps was produced > while training so while training the support sets are unnecessary? and how > the testing reps are generated? > > — > You are receiving this because you commented. > Reply to this email directly, view it on GitHub > <https://github.com/jshtok/RepMet/issues/32#issuecomment-680928642&gt;, or > unsubscribe > <https://github.com/notifications/unsubscribe-auth/ACOBU6X6QF4UQLFYWRELHJTSCUOKTANCNFSM4QL4QK5Q&gt; > . >

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