[Dprglist] How good is computer vision for robotics?
Thalanayar Muthukumar
tnkumar at gmail.com
Thu Oct 7 15:48:19 PDT 2021
I see a healthy skepticism on areas where computer vision may not work - occlusion, poor lighting conditions etc.
Computer vision technology - CNN, pyTorch, RNN coupled with augmentation technologies have improved capabilities a lot in the past few years.
The second aspect to consider is whether basic behaviors using odometry and IMU are good enough for robotics.
I believe it is not an either-or.
When you come closer to the cone in a race, computer vision has an important role to play while earlier on, other techniques may be more beneficial.
I believe the less religious we are and understand the right mix of technologies in its context, the better and more fun the solution will be for us.
Many in the DPRG are eagerly awaiting their OAK 😀 and we will all have fun.
Regards.
- Kumar
Sent from my iPhone
> On Oct 7, 2021, at 3:50 PM, dprglist-request at lists.dprg.org wrote:
>
> Send DPRGlist mailing list submissions to
> dprglist at lists.dprg.org
>
> To subscribe or unsubscribe via the World Wide Web, visit
> http://lists.dprg.org/listinfo.cgi/dprglist-dprg.org
> or, via email, send a message with subject or body 'help' to
> dprglist-request at lists.dprg.org
>
> You can reach the person managing the list at
> dprglist-owner at lists.dprg.org
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of DPRGlist digest..."
>
>
> Today's Topics:
>
> 1. Re: Would it be possible to make an object detector that
> works on any object? (Murray Altheim)
> 2. Re: Would it be possible to make an object detector that
> works on any object? (Rud Merriam)
> 3. Re: Would it be possible to make an object detector that
> works on any object? (Karim Virani)
>
>
> ----------------------------------------------------------------------
>
> Message: 1
> Date: Thu, 7 Oct 2021 18:08:11 +1300
> From: Murray Altheim <murray18 at altheim.com>
> To: dprglist at lists.dprg.org
> Subject: Re: [Dprglist] Would it be possible to make an object
> detector that works on any object?
> Message-ID: <01f62af4-b155-08be-f608-9cd06fc819cd at altheim.com>
> Content-Type: text/plain; charset=utf-8; format=flowed
>
>> On 7/10/21 11:16 am, Karim Virani via DPRGlist wrote:
>> In some ways yes and in some ways no. At this very moment I'm pursuing
>> the "solve" through classic vision processing algorithms. Meaning [...]
>
> Hi Karim,
>
> I'm looking forward to playing with the OAK-D-Lite when mine arrives in
> the mail. But the hoo-hah is a bit over the top.
>
> One of the things I tend to go back to on a lot of the AI/Machine Learning
> technologies are arguments by Rodney Brooks on the real efficacy as well
> as failures of MIT's Shakey robot, which also reminds me of David
> Anderson's idea of having a robot that can run around in my living room.
>
> So it's all fine and dandy if a robot vision system can work on a flat
> grey floor with Block World in primary colors, but what about my actual
> living room, which has a Persian rug with lots of patterns on it and a
> sometimes-moving cat obstacle? And believe me, my cat is not very cat-
> shaped when he's not moving. Sure, a vision system will be able to
> differentiate objects on a plain background that don't overlap, but can
> it locate a mobile phone sitting on a book, sitting on my rug? Even if
> the lights in my lounge are on low, or I'm using the color-control on
> the lights ("romantic mood"!) so the rug changes color?
>
> I agree that we should hear more about the failures, about edge cases,
> rather than all the boosterism which does the industry no favours. Let's
> just say my own experience in the field has left me with a very healthy
> skepticism of its claims.
>
> And looking back at the world of AI for the past half century or so it
> hardly seems that's going to change any time soon. Flying intelligent
> autonomous talking cars that can vacuum your brain clean and discuss the
> problems with your mother-in-law whilst simultaneously making you wealthy
> by creating Elon Musk memes on YouTube and mining cryptocurrencies as a
> background task, that'll happen next year. On Mars.
>
> Cheers,
>
> Murray
>
> ...........................................................................
> Murray Altheim <murray18 at altheim dot com> = = ===
> http://www.altheim.com/murray/ === ===
> = = ===
> In the evening
> The rice leaves in the garden
> Rustle in the autumn wind
> That blows through my reed hut.
> -- Minamoto no Tsunenobu
>
>
>
> ------------------------------
>
> Message: 2
> Date: Thu, 7 Oct 2021 01:10:36 -0500
> From: Rud Merriam <rudmerriam at gmail.com>
> To: dprglist at lists.dprg.org
> Subject: Re: [Dprglist] Would it be possible to make an object
> detector that works on any object?
> Message-ID: <83ad10ac-48b9-08b0-4e49-49c79a31e615 at gmail.com>
> Content-Type: text/plain; charset="utf-8"; Format="flowed"
>
> I have one ordered also. Should get here in the December shipments.
> We'll all have to share our experiences.
>
>
> -73 -
> *Rud Merriam K5RUD*
> /Mystic Lake Software/ <http://mysticlakesoftware.com/>
>
>> On 10/7/21 12:08 AM, Murray Altheim via DPRGlist wrote:
>>> On 7/10/21 11:16 am, Karim Virani via DPRGlist wrote:
>>> In some ways yes and in some ways no. At this very moment I'm pursuing
>>> the "solve" through classic vision processing algorithms. Meaning [...]
>>
>> Hi Karim,
>>
>> I'm looking forward to playing with the OAK-D-Lite when mine arrives in
>> the mail. But the hoo-hah is a bit over the top.
>>
>>
>> Murray
>>
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL: <http://lists.dprg.org/pipermail/dprglist-dprg.org/attachments/20211007/b43aee41/attachment-0001.html>
>
> ------------------------------
>
> Message: 3
> Date: Thu, 7 Oct 2021 13:53:25 -0500
> From: Karim Virani <pondersome64 at gmail.com>
> To: rud.merriam at gmail.com
> Cc: DPRG <dprglist at lists.dprg.org>, Rud Merriam <rudmerriam at gmail.com>
> Subject: Re: [Dprglist] Would it be possible to make an object
> detector that works on any object?
> Message-ID:
> <CAKtnkiwhguZ_ORr5VNMn-Si1su++NfazU9iQaqU+ZDjM8YSkmg at mail.gmail.com>
> Content-Type: text/plain; charset="utf-8"
>
> How about a D-Lite-full December DRPG virtual unboxing party?
>
> On Thu, Oct 7, 2021 at 1:10 AM Rud Merriam via DPRGlist <
> dprglist at lists.dprg.org> wrote:
>
>> I have one ordered also. Should get here in the December shipments. We'll
>> all have to share our experiences.
>>
>> -73 -
>> *Rud Merriam K5RUD*
>> *Mystic Lake Software* <http://mysticlakesoftware.com/>
>>
>> On 10/7/21 12:08 AM, Murray Altheim via DPRGlist wrote:
>>
>> On 7/10/21 11:16 am, Karim Virani via DPRGlist wrote:
>>
>> In some ways yes and in some ways no. At this very moment I'm pursuing
>> the "solve" through classic vision processing algorithms. Meaning [...]
>>
>>
>> Hi Karim,
>>
>> I'm looking forward to playing with the OAK-D-Lite when mine arrives in
>> the mail. But the hoo-hah is a bit over the top.
>>
>>
>> Murray
>>
>> _______________________________________________
>> DPRGlist mailing list
>> DPRGlist at lists.dprg.org
>> http://lists.dprg.org/listinfo.cgi/dprglist-dprg.org
>>
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL: <http://lists.dprg.org/pipermail/dprglist-dprg.org/attachments/20211007/4e80488a/attachment-0001.html>
>
> ------------------------------
>
> Subject: Digest Footer
>
> _______________________________________________
> DPRGlist mailing list
> DPRGlist at lists.dprg.org
> http://lists.dprg.org/listinfo.cgi/dprglist-dprg.org
>
>
> ------------------------------
>
> End of DPRGlist Digest, Vol 61, Issue 7
> ***************************************
More information about the DPRGlist
mailing list