For my project I need to replace the OV2640 camera with OV3660 as the OV2640 does not recover if put to sleep. The OV3660 does recover and is a better camera.
I can buy the stock Xiao ESP32S3 Senses, throw away the OV2640 cameras and buy OV3660 cameras. However, that’s wasteful. Is there a way to obtain Xiao ESP32S3 Senses with OV3660 or alternatively can I buy just the base ESP32S3 and buy or make the bare expansion board and source OV3660 cameras separately?
Surely a Xiao ESP32S3 Sense with OV3660 for say $1 more would be a good product for SeeedStudio to offer?
You can purchase the Xiao ESP32S3 without the camera and design a custom breakout or expansion board to interface with the OV3660. If using detachable OV3660 modules, include a connector (e.g., FPC or JST) on your expansion board.
The OV3660 is pin compatible with the OV2640. I just need the existing expansion board really but without the broken camera. Hoping someone from Seeedstudio will see this and respond.
I may need 100s of them if my project takes off.
Same here. I have some with 75mm ribbon cables. There are 21mm ones available on AliExpress but they are a lot more expensive. I paid just under £2 for my 75mm ones. The longer ribbon necessitated a change to my 3D-printed support bracket to allow it to loop out of the way but otherwise was not a problem. It’s a shame the 2640 is supplied as standard. It really undermines the assertion that the ESP32S3 Sense is a viable device for battery operation.
I see SeeedStudio now supply the OV3660 camera as standard. Would have been nice to have had some warning before I went and bought 20x the old product.
Xiao is rather expensive. I need another 20x just to keep my field testing work going but probably going to switch to using a generic ESP32S3 module with the same camera interface connector. I also need to be able to focus the cameras to 115mm and the standard ones come focused to infinity and superglued. I’ve found a Chinese supplier who can do OV3660 WITHOUT the superglue if I buy 500 so I’m going to do that.
WoW, that’s a serious undertaking. Did you try to contact Seeed , see if they can accommodate you? they have auto focus unit AFAIK too?
sounds like a cool project too
Idunno if it matters but I believe the frame rate to be lower on the ESPCam than on say a GroveVision Ai2 or S3 cam. but you probably know a lot more on the subject, I’m a casual user of the camera stuff …LOL (messing with auto focus)
It might be worth a go I guess.
Autofocus is not required for this. It’s always going to be fixed focus once setup.
If you want to know what the project is about it’s described here: https://www.buzzcopper.org
I believe you have a few Asian Hornet incursions over in the USA but it’s probably not such a threat to you guys as it is in the UK and it is an every day nightmare in most of Northern Europe now.
WOW-SA! Asian Hornet Alert!
Yes, that is really, really great … I have seen those bastids destroy bee’s the Hives and some texas areas, have big swarms of those destructive wasps too…
So I think you could do well to team-up or partner up with seeed and the project portal, GREAT application bring them along they can probably discount some for project submissions. Worth a look/email…
I see some Vision AI in there too maybe?
Really good stuff you are doing.
Yup. I think it’s a worthwhile project and the beekeepers I’ve approached think it’s great too. Hence my urgent need for 20x more processors just to meet demand while I’m field testing.
I’m not actually doing any AI on the Xiao. I’ve written my own more primitive Machine Vision algorithm that simply looks for collections of contiguous darker pixels that are roughly the right size and which were not in the previous picture. The AI happens up in the Cloud. Currently using Google Vision but it’s not been trained on my images so doesn’t work very well. We’re going to create our own cloud-based AI engine for inferencing when we have enough training data.
I will send Kevin an email as you suggest just sharing what I’m doing and offering the chance for SeeedStudio to stay involved but I have two of the generic modules arriving next week from UK and 10 more on order from China so I can do some proper testing before I redesign the plastics and place a bulk order. I have some concerns about the power management on those generic boards: it must be able to run off an 18650 battery for a week. So maybe there’s still an opportunity for Xiao. We’ll see.
regards
Jon
Excellent John… It really is. So I would say you could use the senscraft AI , Grove_vision AI2 and camera the training is the key and pretty easy.
(as an aside) I read once about cutting our grass too short for the pollinators of the fruit trees. I was skeptical at first… Everyone in our area cuts there grass so low it almost burns from the sun. I stopped doing that and let it get longer 2x longer and some small flowers and other weeds came in(lots more bees). But what a difference in our fruit production…WOW I was amazed and Now I’m a disciple. It has a HUGE effect on our Fruit trees. I tell everyone that asks how I get so many and some very Large fruits too(because they are set earlier and on there longer) So I’m paying attention.
I can see some major points:
Power consumption
Size
data collection
identification accuracy
I would think top concerns
please keep posting and updating a lot of folks are on vacation but I’m confident the views would pickup as members, viewers return.
IMHO, This fits in well with Seeed Studios Earth care approach and teaching tech, you would be hard pressed to find a better cause and effect example so well matched with a tech solution.
they say that " Necessity is the mother of invention " so saving the planet one idea at a time
sent you a message and I linked him in to the discussion…
You’ll hear something (response time may be indirectly proportional to Work Load environment)
HTH
GL PJ
My pops used to sing a song about them at dinner " Mary went a walking, Walking thru the grass, Along came a bumble bee and stung her in the AZ I said before beware of bees… " LOL
Buy base Xiao ESP32S3 and design a tiny camera board with OV3660. You can order a small run of PCBs and assemble 5–10 units with minimal cost and no waste.
PJ 
(
