I have a couple of security cameras outside and the motion and people detection on them is terrible. Many, many false alarms.
I’m working on an app that will pass their jpg files through Vizy so it only alerts me when there’s something of real interest in the images.
I got the detection working this morning (with your hint). Next I have to work on the GUI. The Pet Companion tutorial should help with that.
Hi Edward,
Any news on the European species data model?
Thanks!
Dave
Hi Dave,
We’re working on it. 
We’ve determined which network to use (Efficientdet running via Tensorflow 2). We hope to have it released before August. We’ll provide a European bird detection network (in addition to the North American), but as mentioned before, it should also be fairly straightforward to create your own networks, Sorry this isn’t available now.
Edward
I bought this specifally to create custom detection/classifications. Marketing made me think that plug and play and modify was already baked in.
I tried the from scratch thing in the past and got bogged down in getting the hardware and low level software/libraries up and running. Thought this would let me focus on learning the vision side of things.
For now, without any documentation/tutorials on taking something like the birdfeeder example and training for a different critter this is pretty useless to me. I was expecting an out of box development experience but I guess that was expecting too much too soon.
I will just keep checking back here periodically. Fortunately, I have a hundred other projects that need tending to.
Even just stepping us through the process of creating something like the birdfeeder project so that I can get my head around the steps involved in the developing and deploying an app on this hardware would be very helpful.
Hello,
Have you seen this page?
https://docs.vizycam.com/doku.php?id=wiki:pet_companion2
It’s a guide for creating your own Vizy application and is reasonably easy to follow.
The training of your own network is something we’re targeting for release before September. Image detection/classification CNNs are relatively easy to use (just feed them pictures and get the results – this is what we have now), but building your own CNN detector or classifier is a different level of complexity. We created the birdfeeder network on an AWS machine with special graphics assist for AI. Fortunately, things are much easier now, but it’s taking us some time to set up a reasonably easy path for customers like yourself to follow, so you can create your own network from a set of training images and expect decent results (good speed/accuracy).
If there were expectations that were oversold in the marketing info, we will correct (let us know). Our goal is to provide something that makes these kinds of vision applications much easier to use, deploy, and author. It’s a task that’s taking more time than we originally thought. We have achieved much of that goal with the current software, but there’s still much work to do with image detection/classification (as described) and API documentation – for this we apologize. We will get there though.
Edward
.
As a developer myself, the statement “taking more time than we orginially thought” should be on a T-shirt. 
Having attempted to dig into vision off an on in the past, the task you have set before you of making this stuff accessible is a big one. I realize you have to get the infrastructure figured out and in place before you can really focus on simplifying the user experience.
Based on what your PixyCam did for the FIRST community in making vision more of a commodity and not just for the elite teams I have faith that you will be successful in this next big jump up.
I will wait patiently on you guys and hack at the examples in the meantime to get some level of understanding with some of the basics.
I would like to start by hacking the bird feeder defender to defend my strawberry patch from rabbits and birds. Any tips/suggestions? I did’t see rabbits in any of the classifications (not a big bird feeder nuisance I know). Also, in this implementation I don’t care about bird species so would it be better to use somthing like the Coco library? I didn’t see rabbits in Coco either though.
Thanks for the hard work you’ve done and the hard work to come.
This is really great – you’re not afraid to get your hands a bit dirty and you have a cool application that you want to create. Coco isn’t going to give you what you want, unfortunately. There’s a really good image resource of labeled images provided by iNaturalist. It has hundreds of labeled images of each common plant and animal species. With these images and a decent guide, you can create your own detection networks. We took images from the iNaturalist database (the 20 most common North American bird species) to create the birdfeeder detection network.
In a future Vizy software release and help from our how-to documentation, you’ll be able to grab the labeled images from this database and create your own CNN for detection/classification of practically anything you want (plant or animal). We’ll also have a way for you train on a cloud-hosted processor.
In the meantime, we’re happy to answer any questions that come up in when writing Vizy code (given the sparse documentation of the vizy/kritter APIs)
Edward
In the meantime starting with the Pet Companion I found it thinks rabbits may be dogs or cats. I have issues with the neighbors cat and dogs so wth bird, cat and dog detection on in COCO I might get a good approximation for V.1 deployment.
Do you guys discuss how to “steer” something like a servo (for garden hose control etc.) based on detected object location in image field?
Really looking forward to the documentation on using iNaturalist to create my own detection/classification and insert them into the system - hurry please!
I did have a go at training for UK birds ( plus a few “pests” ) - see https://github.com/plord12/vizy-training . Seems to work with test images but really waiting for vizicam to move to tensorflow v2 before I get back to this.
Pain point is creating bounding boxes …
iNaturalist has bounding boxes
Perhaps not for 2021. There are lots of labeled images available however.
Well, I’ve not re-checked, but a while back I did find that the common UK birds were not in the 2017 set … pictures only in the 2021 set without bounding boxes. Hence the earlier comment.
We will be moving to a multi-step inference for the Birdfeeder app. A detection network will detect generic birds and pests, and then a classification network will identify the bird species using the detection network’s detection box. This will improve the accuracy, and since the birds are identified with a classification network, bounding boxes aren’t necessary for the training set, just the specific classification (bird species).
Edward
Hows is going ?
I’m followiing along on github and see some great sounding pushes 
Hello,
We are trying to get the next release out, which is probably our biggest release yet. It uses the improved detection and classification networks (Efficientdet and Efficientnet) and has support for “texting” – sending notifications to you phone.
And of course, things are taking longer than expected… We expect the release to be ready mid-September. We have a European bird classification network that works with the new CNNs, and we need some real world testing
Edward
Brilliant … thanks for the update
I’ll be first in line for testing !
Hello,
Just a quickie update – I said mid September on the next release with the European bird classification network (etc). It’s going to take longer 
We’re targeting October 7 for release. Things are progressing, it’s just a complex set of changes including a rewrite of the Birdfeeder app and a new app (Object Detector). Thanks for your patience
Edward
Thanks for the update.
I noticed in the source a new “daytime” detector. Is one of the ideas here to auto turn on/off infrared leds ?
Thanks again.