Go green team! - Githubissues

Arya07 commented 7 years ago

Hey @vvv-school/vvv17-team-green, Welcome everyone to the best team, the green team!!! :D During the pauses between the ML assignments (don't skip them ;)) please start thinking about some demo's ideas! It would be a good idea to have brainstorming at some time, before/during/after dinner. So let's think about some great demo to win!! :1st_place_medal:

towardthesea commented 7 years ago

Let's make the GREEN great again!

jibweb commented 7 years ago

how about giving a "story" around the technical accomplishment ? For example, we could work on the "green" robot, which would properly recycle, putting the trash in the appropriate bin. Quite similar idea from a technical standpoint, but with a context for the goal

Arya07 commented 7 years ago

the idea is very good! In our context we have to consider the dataset we have, and how to theach things to the robot

Da: Jean-Baptiste Weibel [notifications@github.com] Inviato: lunedì 6 febbraio 2017 17.59 A: vvv-school/vvv17-demo-team-green Cc: Elisa Maiettini; Author Oggetto: Re: [vvv-school/vvv17-demo-team-green] Go green team! (#1)

how about giving a "story" around the technical accomplishment ? For example, we could work on the "green" robot, which would properly recycle, putting the trash in the appropriate bin. Quite similar idea from a technical standpoint, but with a context for the goal

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<redir.aspx?REF=8TCRBoeLSKcZIaPZaCTcO8YlGXSnpjL6S3LpZoO3Hwa0hwzCsk7UCAFodHRwczovL2dpdGh1Yi5jb20vdnZ2LXNjaG9vbC92dnYxNy1kZW1vLXRlYW0tZ3JlZW4vaXNzdWVzLzEjaXNzdWVjb21tZW50LTI3Nzc0NDAzMQ..>, or mute the thread<redir.aspx?REF=Eb9m_rfJ7EDWZvWviFMkCHlAfuIFZpUc-QKMJditO9W0hwzCsk7UCAFodHRwczovL2dpdGh1Yi5jb20vbm90aWZpY2F0aW9ucy91bnN1YnNjcmliZS1hdXRoL0FEaU5ncUpST2lla19oZEdVOVloWm1NWTBuRVNkWHdsa3M1cloxR0dnYUpwWk00TDROLTU.>.

alexarossi commented 7 years ago

it's a good idea, but it could be tricky for the recognition of the object to recycle. That means either we have very few known objects or it becomes too complicate.

What about:

on the table, there are 3 different bins.
if there is a person in robot's vision range, the robot says something/ greet the person/ whatever
if the person has also another object, the robot asks to put closer the bin to recycle the object. it picks the object and move it in the bin.

On Mon, Feb 6, 2017 at 6:12 PM, Elisa notifications@github.com wrote:

the idea is very good! In our context we have to consider the dataset we have, and how to theach things to the robot

Da: Jean-Baptiste Weibel [notifications@github.com] Inviato: lunedì 6 febbraio 2017 17.59 A: vvv-school/vvv17-demo-team-green Cc: Elisa Maiettini; Author Oggetto: Re: [vvv-school/vvv17-demo-team-green] Go green team! (#1)

how about giving a "story" around the technical accomplishment ? For example, we could work on the "green" robot, which would properly recycle, putting the trash in the appropriate bin. Quite similar idea from a technical standpoint, but with a context for the goal

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<redir.aspx?REF=8TCRBoeL SKcZIaPZaCTcO8YlGXSnpjL6S3LpZoO3Hwa0hwzCsk7UCAFodHRwczovL2dp dGh1Yi5jb20vdnZ2LXNjaG9vbC92dnYxNy1kZW1vLXRlYW0tZ3JlZW4vaXNz dWVzLzEjaXNzdWVjb21tZW50LTI3Nzc0NDAzMQ..>, or mute the thread<redir.aspx?REF=Eb9m_rfJ7EDWZvWviFMkCHlAfuIFZpUc-QKMJd itO9W0hwzCsk7UCAFodHRwczovL2dpdGh1Yi5jb20vbm90aWZpY2F0aW9ucy 91bnN1YnNjcmliZS1hdXRoL0FEaU5ncUpST2lla19oZEdVOVloWm1NWTBuRV NkWHdsa3M1cloxR0dnYUpwWk00TDROLTU.>.

— You are receiving this because you are on a team that was mentioned. Reply to this email directly, view it on GitHub https://github.com/vvv-school/vvv17-demo-team-green/issues/1#issuecomment-277747674, or mute the thread https://github.com/notifications/unsubscribe-auth/AYLSdinKOsDtij9o6tTY0u-7ZPOtmnv5ks5rZ1RhgaJpZM4L4N-5 .

damianomal commented 7 years ago

Very nice ideas! :)

This idea of a "green robot" is really nice, we can talk about it extensively in a few minutes, we have to ask ourselves a lot of questions, for example "how easy would it be for us to find three different bins, differently colored, in the short time we have?".

By the way, as I said, I really liked these suggestions, and I really can't wait to work on this with all of you!

jibweb commented 7 years ago

It's more about the scenario than about the details in this case, we can use any bin, or not even a bin, just a place on the table using a hard-coded position, etc. For the recognition, unless we find a suitable dataset to finetune on, we can simply work on a subset of imagenet, like can, bottle and a few others It's more about the proof of concept than a fancy consumer ready product, so it can have a few rough edges.

Also, making the robot speak without speech recognition is not a great idea interaction-wise

Arya07 commented 7 years ago

I agree with you @jibweb but we need a way to interact with the user, eg. to ask to put an object nearer or something like that

Arya07 commented 7 years ago

during the meeting after dinner with the teachers we can ask what possibilities we have with speech

damianomal commented 7 years ago

I also agree with @jibweb and @Arya07, the robot could just recognize a really small subset of objects and an example of interaction could be:

the robot recognizes one of the objects on the table
it asks the user to put on the table the corresponding bin (plastic, paper, etc.)
it grabs the object and puts it in the bin
asks the user to take the bin away from the table
and repeats until the table is empty

An example of interaction could be a kind of "serious game", we could challenge other users to clean the table with the help of the robot as quick as possible.

alexarossi commented 7 years ago

I agree about the bin/paper/simple position.

Vocal cues and speech recognition are different type of interaction, not sure why it should be not wise. but, we can sure think about something else.

i think that it's okay consider it as proof of concept. but, at this point let's choose only one or two objects. we cannot expand the recognition too much because we cannot spend too much time in looking for a dataset since we already have a lot of things to do.

On Mon, Feb 6, 2017 at 6:56 PM, Jean-Baptiste Weibel < notifications@github.com> wrote:

It's more about the scenario than about the details in this case, we can use any bin, or not even a bin, just a place on the table using a hard-coded position, etc. For the recognition, unless we find a suitable dataset to finetune on, we can simply work on a subset of imagenet, like can, bottle and a few others It's more about the proof of concept than a fancy consumer ready product, so it can have a few rough edges.

Also, making the robot speak without speech recognition is not a great idea interaction-wise

— You are receiving this because you are on a team that was mentioned. Reply to this email directly, view it on GitHub https://github.com/vvv-school/vvv17-demo-team-green/issues/1#issuecomment-277760425, or mute the thread https://github.com/notifications/unsubscribe-auth/AYLSdjEUB9wrFLnTKKa1d3bcUMc36eVGks5rZ17WgaJpZM4L4N-5 .

anatanevska commented 7 years ago

@Arya07 @jibweb even without speech recognition, it'd be possible to see whether the user has helped the robot or not, e.g. - if the robot asks the user to put the object closer to him and then measures the distance again, it can see whether the user helped him or not.

alexarossi commented 7 years ago

i think we don't need to recognize the object. just recognize the presence of the object.

On Mon, Feb 6, 2017 at 7:04 PM, Damiano Malafronte <notifications@github.com

wrote:

I also agree with @jibweb https://github.com/jibweb and @Arya07 https://github.com/Arya07, the robot could just recognize a really small subset of objects and an example of interaction could be:

the robot recognizes one of the objects on the table

it asks the user to put on the table the corresponding bin (plastic, paper, etc.)

it grabs the object and puts it in the bin

asks the user to take the bin away from the table

and repeats until the table is empty

An example of interaction could be a kind of "serious game", we could challenge other users to clean the table with the help of the robot as quick as possible.

— You are receiving this because you are on a team that was mentioned. Reply to this email directly, view it on GitHub https://github.com/vvv-school/vvv17-demo-team-green/issues/1#issuecomment-277762996, or mute the thread https://github.com/notifications/unsubscribe-auth/AYLSdrZg2UDoXmuzGr3IAMQzr-lWOO8eks5rZ2DKgaJpZM4L4N-5 .

damianomal commented 7 years ago

Sure! :)

I meant in the case we want to "differentiate" the different types of waste.

Arya07 commented 7 years ago

maybe we can just distinguish if an object is one of the object that the robot knows (from the dataset) and so not to put it in the bin or if it is other -> trash

jibweb commented 7 years ago

exactly ! that's also what you're supposed when recycling :-) alsso attached the list of class from imagenet synset_words.txt

jibweb commented 7 years ago

or to make a simpler scenario (that can evolve if we have time), just add one object at a time in front of the robot, and put it in a pile, either the left or the right if it can recycle it or not

alexarossi commented 7 years ago

yeah! I like this. :)

Arya07 commented 7 years ago

meeting upstairs in 5 minutes!! :)

Arya07 commented 7 years ago

Lawrence where are you?:)

vvv-school / vvv17-demo-team-green

Go green team! #1