{ "id": 5195, "title": "Fixed 10 gestures from test video sequence?", "url": "/competitions/multi-modal-gesture-recognition/discussion/5195", "author_name": "", "post_date": "2013-07-24T18:14:25.527Z", "votes": null, "comment_count": 4, "views": 1389, "content": "

A Quote from Miguel:

\n

"Dear Brian,

\n


The test dataset will be composed by sequences of one minute, which will be contained 10 gestures from the vocabulary (without replications). People who perform the gestures are the same as training and evaluation dataset. It is not necessary to segment the gesture in frames, only label 10 gestures in each sequence. Feel free to ask any question.

\n

Greetings,

\n


Miguel Reyes Estany"

\n

 

\n

My question is:

\n

"Does that mean for testing dataset, the label number is fixed at 10 unique numbers (out of 20). That's quite different from validation dataset ( I assume there will be around 10 gestures (NOT fixed) in one sequence and there might be several same gestures in a video sequence). Can you give us some further instructions concerning with the validation dataset since validation set and test set have different structure and it will cause the design of our system to be quite different for validation set and testing set.

\n

Thank you

\n

 

", "messages": [ { "id": "27613", "postDate": "07/24/2013 18:14:25", "content": "

A Quote from Miguel:

\n

"Dear Brian,

\n


The test dataset will be composed by sequences of one minute, which will be contained 10 gestures from the vocabulary (without replications). People who perform the gestures are the same as training and evaluation dataset. It is not necessary to segment the gesture in frames, only label 10 gestures in each sequence. Feel free to ask any question.

\n

Greetings,

\n


Miguel Reyes Estany"

\n

 

\n

My question is:

\n

"Does that mean for testing dataset, the label number is fixed at 10 unique numbers (out of 20). That's quite different from validation dataset ( I assume there will be around 10 gestures (NOT fixed) in one sequence and there might be several same gestures in a video sequence). Can you give us some further instructions concerning with the validation dataset since validation set and test set have different structure and it will cause the design of our system to be quite different for validation set and testing set.

\n

Thank you

\n

 

", "rawMarkdown": "", "votes": null }, { "id": "27660", "postDate": "07/25/2013 19:18:38", "content": "

Dear Steven,

\n

 

\n

I want to clarify that the final dataset is composed of several sequences of one minute approximately, which each contain ten gestures (out of 20). As much as the people who make the gestures and the data structure are exactly the same.

\n

I hope this does not change much your designs, sorry for the incovenience. Feel free to ask any question.

\n

Greetings,

\n

 

\n

Miguel Reyes Estany

", "rawMarkdown": "", "votes": null }, { "id": "27691", "postDate": "07/26/2013 15:44:25", "content": "

Sorry, just to try to get this 100% clear:  In the evaluation set, each video contains *exactly* 10 true gestures (plus some fake stuff).  Never 9, never 11, right?

\n

 

\n

Are these 10 gestures always unique, or is it allowed to have repeats?

", "rawMarkdown": "", "votes": null }, { "id": "27825", "postDate": "07/30/2013 08:50:57", "content": "

Dear Brian,

\n

 

\n

There will be 10 unique gestures (out of 20) in each sequence, without repetitions (plus some fake stuff).

\n

Greetings.

", "rawMarkdown": "", "votes": null }, { "id": "27836", "postDate": "07/30/2013 14:32:36", "content": "

We are finalizing the creation of the test dataset, and I would like to clarify that each sequence will contain between 9 and 12 gestures (an average of ten), without repetitions and with some noisy gesture.

\n

Greetings,

\n

 

\n

Miguel Reyes Estany

", "rawMarkdown": "", "votes": null } ], "comments": [ { "id": 27660, "author_name": "miguelreyesestany", "author_url": "", "post_date": "07/25/2013 19:18:38", "content": "

Dear Steven,

\n

 

\n

I want to clarify that the final dataset is composed of several sequences of one minute approximately, which each contain ten gestures (out of 20). As much as the people who make the gestures and the data structure are exactly the same.

\n

I hope this does not change much your designs, sorry for the incovenience. Feel free to ask any question.

\n

Greetings,

\n

 

\n

Miguel Reyes Estany

", "votes": null, "replies": [] }, { "id": 27691, "author_name": "brianfeldstein", "author_url": "", "post_date": "07/26/2013 15:44:25", "content": "

Sorry, just to try to get this 100% clear:  In the evaluation set, each video contains *exactly* 10 true gestures (plus some fake stuff).  Never 9, never 11, right?

\n

 

\n

Are these 10 gestures always unique, or is it allowed to have repeats?

", "votes": null, "replies": [] }, { "id": 27825, "author_name": "miguelreyesestany", "author_url": "", "post_date": "07/30/2013 08:50:57", "content": "

Dear Brian,

\n

 

\n

There will be 10 unique gestures (out of 20) in each sequence, without repetitions (plus some fake stuff).

\n

Greetings.

", "votes": null, "replies": [] }, { "id": 27836, "author_name": "miguelreyesestany", "author_url": "", "post_date": "07/30/2013 14:32:36", "content": "

We are finalizing the creation of the test dataset, and I would like to clarify that each sequence will contain between 9 and 12 gestures (an average of ten), without repetitions and with some noisy gesture.

\n

Greetings,

\n

 

\n

Miguel Reyes Estany

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "27613": "", "27660": "", "27691": "", "27825": "", "27836": "" }, "source": "meta" }