[somewhatshocked]

Okay. It's nerd time.

Do you like tinkering with Raspberry Pi stuff? Then this may be for you.

From the opening paragraph:

This tutorial was inspired by the Entaniya waterproof case for the Raspberry Pi Camera Module. This case protects the Camera Module without distorting the image. With it, we can submerge a camera into the depths of an aquarium and keep an eye on our fishy pals by streaming video. If that doesn’t appeal, you can adapt this tutorial to be able to monitor a video stream of anything you like, whether it be indoors or outdoors.

Looking forward to seeing what y'all do.

 

[HillstreamBiotoper]

I do not currently know too much about Raspberry Pi. Not too long ago, I got into Python because my friends were doing all kinds of fun stuff with it and similar things (they always play with microcontrollers and microcomputers). I am just about a fourteenth of the way to knowing it. Anyway, could you get the Raspberry Pi to use recognition with individual fish/shrimp (ones with more differentiated patterns/shapes) and it alert you if there is one that has gotten a clear injury (seen by discoloration of certain parts of the body)?

 

[somewhatshocked]

Anyway, could you get the Raspberry Pi to use recognition with individual fish/shrimp (ones with more differentiated patterns/shapes) and it alert you if there is one that has gotten a clear injury (seen by discoloration of certain parts of the body)?

Absolutely. While it would probably take a lot more effort to get a functional system to spot injuries (for obvious reasons), there are probably already programs available that can be modified to help identify individual fish or types of fish. There are a few different programs people use to identify packages, humans, pets and all that. Even some that can identify different humans.

My favorite thing about the Pi system, since devices vary from $5 to $100 or so, is the accessibility. Nearly everyone can afford to tinker with the platform. Some people use them to fully automate their tanks, homes, you name it. At this point in the nerd realm, you don't even really need to be well-versed in coding in order to do a lot. Every RasPi runs a full graphical desktop/graphical user interface - a modern linux desktop.

I've got an 8gig RasPi connected to my TV that I use all the time for basic web browsing. But it also serves as a spot to monitor several cameras, tracks temperature and humidity in every room, controls the drip lines for my container plants, I video conference with it occasionally, it monitors my networked storage, tells me when I need to refill my bird feeder, even controls one of my old CurrentUSA LED fixtures with a cheap IR emitter. All kinds of pointless crap, honestly.

If you're a dork like me, it's a no-brainer. Especially if you want to use one to do tank stuff.

 

[minorhero]

I do not currently know too much about Raspberry Pi. Not too long ago, I got into Python because my friends were doing all kinds of fun stuff with it and similar things (they always play with microcontrollers and microcomputers). I am just about a fourteenth of the way to knowing it. Anyway, could you get the Raspberry Pi to use recognition with individual fish/shrimp (ones with more differentiated patterns/shapes) and it alert you if there is one that has gotten a clear injury (seen by discoloration of certain parts of the body)?

To my knowledge this program does not currently exist. So if you wanted to track fish injury, you would need to code it yourself. There was a LOT of talk about the FELIX system having this functionality (FELIX being an aquarium controller similar to apex) and they did a good amount of marketing telling people they were going to have injury tracking for fish, but that feature ended up being vaporware.

 

[somewhatshocked]

So if you wanted to track fish injury, you would need to code it yourself.

I'm not even sure it could be done well - yet. I think injuries would have to be pretty substantial even with skilled developers at the helm.

The only thing I've seen so far that works well is species identification.

 

[reverse.engineer]

Building a smart tank and components like smart monitoring and stuff has something I've recently taken a bit of interest in. I didn't know there was a raspi camera module you could add to your tank!

I'm not even sure it could be done well - yet. I think injuries would have to be pretty substantial even with skilled developers at the helm.

The only thing I've seen so far that works well is species identification.

Wouldn't it be possible to train a neural network on things like external signs of infections and injuries (e.g missing scales or ich white spots) and then feed it with images of the fish to classify whether or not the fish is potentially injured or sick?

I think that the code to build algorithms that detect things like this using images is likely here. It's just a matter of using and training the right algorithm for the job on the data it would need.

Sent from my SM-F721U1 using Tapatalk

 

[somewhatshocked]

I think that the code to build algorithms that detect things like this using images is likely here. It's just a matter of using and training the right algorithm for the job on the data it would need.

It would be pretty substantial software for a Raspberry Pi. Would depend upon the specific Pi one is using, the processor, RAM, the camera or cameras involved (there are several Pi cams), et al.

But there are existing projects that aren't far off from identifying species. Some can identify individual humans and dogs.

 

[somewhatshocked]

Raspberry Pi just released a global shutter camera that will make this sort of thing even more possible, @HillstreamBiotoper. cc @reverse.engineer.