Collect new dataset for human action recognition and motion prediction

[kouroshD]

In this issue, we track the activity for the new dataset we would like to collect. We will use iFeel shoes and Xsens.

We should decide on the set of actions and the context we want to collect the data. Here is a suggestion: Context: daily life activity in the lab

Actions:

• Walking
• Standing
• Rotating
• Sitting on a chair
• Greeting (saying hello by hand)
• maybe opening doors

I want to collect about 20 mins of data, one shot.

What do you say @DanielePucci about the context of the activities and the set of actions? Do you have other suggestions. For this https://github.com/ami-iit/element_human-action-intention-recognition/issues/64 , maybe we even can use the data we used for workshop and we do not collect new data. What is your suggestion in general? Should we collect this new data, or we go with the data we have and later we will see if it is necessary more data and action set.

[DanielePucci]

I think that the above looks fine. Probably, also @paolo-viceconte can help you since today at the vertical he just told me that he wants to take similar data! Join forces guys!

CC @S-Dafarra @raffaello-camoriano

[paolo-viceconte]

I think that the above looks fine. Probably, also @paolo-viceconte can help you since today at the vertical he just told me that he wants to take similar data! Join forces guys!

Yes, another dataset collection is planned for the ADHERENT pipeline as well. We still have to plan it in detail, but it could make sense to try with some actions in addition to locomotion (such as sitting or simpler ones) and/or with some perception in the loop (for instance, crouching to walk with low ceiling or side-walking to cross narrow passages).

I don't know which time horizon you have in mind for this dataset collection @kouroshD but we can definitely help each other :)