Hacker News Reader

Comment by mrafiee | Hacker News Reader

Are you suggesting not using cloud for privacy concerns? Based on the feedback from people's comments here, I realize we should do more to alleviate the privacy concerns. Curious on to know your thoughts about the following aspect: As I mentioned, we only store short video clips corresponding to events that the user created (we already discard the other motion clips that are deemed as irrelevant by our models.) We also allow users to delete the the alerts they have received and when they delete each alert, we permanently delete the corresponding video clips from our dbs... would that alleviate your privacy concerns?

We actually built our first prototype using RPi, we tried 3-4 different RPi cams, the image quality of all of them was very poor. Also the final cost would much higher than the cameras we are using right now...

Supporting IFTTT is in our near term road map. Appreciate the suggestions!

View on HN

It was more that if my internet connection goes down, are the cameras useless? What if the internet/AWS is just "slow" one day - will the notifications be delayed significantly making any "reactive" integrations pointless/ludicrously delayed? If things can run locally (doing inference for multiple cameras via a single "box" you plug in to your WiFi router etc) then you can be super-fast with IFTTT integrations.

My main line if thought was that I built basically your product for spotting when a cat climbed into my plant pots using ML and a RPi3 - the idea was that when it saw the cat, it would squirt a water pistol at it to scare it away - inference on the RPi 3 was too slow (if I was doing this now I'd use a coral accelerator maybe) and by the time it realised a cat had got into the plant pots, the cat had already taken a shit and left. I worry that your product might suffer from similar end to end latency. Niche use-case? Perhaps. I have Amazon Blink cameras here and the IFTTT integration is delayed by about 30 seconds so by the time you get a notification there is someone at your door it is to late to do anything as they will.have already left/kicked the door in by then etc. Doing all this locally would be super fast

My main concern was not really about privacy - you'll need to cover GRPR if someone from the EU happens to walk into frame of one of your customers' cameras one day in the future anyway (Good luck)

Love that use case :) re your point about latency, that is one of the main reasons we are doing all the inference on the cloud. Almost all deep learning models (at least CV models) need GPU to run with low latency. That's true that if you have set up an automated response from another device, it may still work if the internet is down but for alerting the user, you would still need internet connection even with a central hub...

FTR, a simple RPi4 yields ~4FPS on a DL object detection model. You may need cloud for your central hub, but neural network inference can be done locally.

By 4FPS do you mean 250ms latency per frame? What OD model specifically did you get that latency from on RPi4?

It was my understanding that GP's main concern was the service shutting down should his internet fail, and not one of privacy.

I see. That is actually a great point and we have heard that concern from others as well. The cameras we are offering now come with local storage. We have been thinking of adding some capabilities when camera is offline but since we do the inference on the cloud, the smart alert and video indexing features (our main value props) would not work offline.

Than what is your selling point against all those Chinese cloud based services?

Our main value prop is more powerful event detection capabilities that what anyone else is offering (and at a lower price.)

There will be many kinds of customers, but among them those who value privacy and those who want convenience. One of them will drive the bulk of sales, the other maybe not. On the other hand, if this is a value prop that excites people you’ll surely have the incumbents consider the economic threat and may add that feature as needed.

Ya I think finding the right balance between privacy related risks and cost/convenience/features is a an important aspect of this space.

Not OP, but if you want privacy conscious users to use your business, they will want to be able to host it themselves.

Ya I see your point. We have tried to take the privacy seriously from the beginning (and now realize we need to do more) but as far as not using the cloud at all, that would be a significant limitation on the type of inference that can be done locally and I think the benefits may outweigh the associated risks for many users/use cases...

I have a number of cloud enabled cameras now, but I’m also building my own system to keep everything local-only.

If you want to sell me an ML-based system, you’re welcome to train the models in the cloud, but they have to run on local-only assets. And you have to give me complete control over downloading new models periodically to a machine of my choice, and the updating my local devices.

Thanks for the feedback. What are you using the cameras for?

Even storing short video clips in the cloud is not good enough.

Even storing short video clips in the cloud is simply not good enough.

How about giving the users full control over them and let them delete them individually and also all of the past alerts at once after they view them?

Reply on news.ycombinator.com