how to calculate classification score in non-key frame?

[bxcxa]

hi,
paper says:“For each sequence set Sk , We have its corresponding mask M(Sk) and classification score C(Sk)”. But I don't find anywhere describing how to calculate classification in the non-key frame.

The framework of seq mask r-cnn only calculate the mask of non-key frame ，not cls.

Thank you!

[HuaijiaLin]

Thanks for your question on details. It seems we do not describe how to obtain the classification score for the non-key frames in the paper.

In the implementation, for a non-key frame, with its mask given, we first calculate the bounding box of this mask and then apply the classification head (of seq mask r-cnn) to obtain the score for this bbox.

The function seq_scoring describes this procedure.

[bxcxa]

get it,thank you  for detailed explanation.

…

---Original--- From: ***@***.***> Date: Tue, Mar 15, 2022 00:49 AM To: ***@***.***>; Cc: ***@***.******@***.***>; Subject: Re: [dvlab-research/ProposeReduce] how to calculate classificationscore in non-key frame? (Issue #5) Thanks for your question on details. It seems we do not describe how to obtain the classification score for the non-key frames in the paper. In the implementation, for a non-key frame, with its mask given, we first calculate the bounding box of this mask and then apply the classification head (of seq mask r-cnn) to obtain the score for this bbox. The function seq_scoring describes this procedure. — Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android. You are receiving this because you authored the thread.Message ID: ***@***.***>