What we call smart cameras are essentially connected cameras because what they do is stream video over the Internet. These gadgets perform no analysis on the data; the connectivity serves only as a conduit for video transfer.
The connected camera is a ubiquitous device that has various applications from homes, streets, stores, offices and even on peoples’ bodies. A study showed that crime rates drop at places with cameras and a small gesture like placing a dummy camera seems to do the job.
The concept of smart home application for connected cameras is accessible to most people. It is fascinating because it creates a way to see into your home regardless of where you are, thereby giving users that extra peace of mind. However, all the connected cameras on the market are designed to work on a motion detection principle, meaning if they sense motion they send you an alert regardless of what was seen. This design principle is a big problem for consumers because it creates an avalanche of meaningless data (like those 28 videos of your cousin John walking around) and these companies charge users exorbitant storage fees for these videos.
Deep learning, computer vision and face recognition have been extensively deployed in various image analysis applications and are no-brainers when it comes to integration with connected cameras. However, given its complexity, it is a difficult problem to solve which is why there has been a lot of talk with different companies releasing their impractical application of this technology that users don’t understand and can’t use. One thing the former CEO of Nest (Tony Fadell) was right about is in keeping smart home simple and not complicated.
For example, Nest just released a Nest IQ camera that deploys face recognition for their users. It is supposed to recognize when “familiar faces” get captured by the camera, which means it isn’t user-defined and the camera will have to see a face first. If you are lucky to get a notification, you have to go to the familiar face section, which contains a multitude of faces that have been seen by the cameras making that another outlet for data overload.
Due to this impractical approach and poor implementation the application is riddled with errors such as high false positives/negatives that can happen when the camera is facing TV, mirrors or even portraits. This drives users to frustration and eventually turning off their cameras. In addition to these flaws, it would also cost users more money to use the advanced Nest aware feature because it is “intelligent”. The sad thing about this is that like a true herd, other companies in the space have the same version of this implementation in exactly the same way, which means that all of them essentially have the same flaws in the design.
At CleepCam we are breaking with this monotonous tradition. We have developed a machine learning enabled camera with a user-defined platform, which enables users to easily teach their cameras who they want it to recognize by simply uploading pictures to an image gallery. Rather than “familiar faces” users would have a friendly and practical image gallery interface that users can fully control.
The concept is about turning an image gallery into an intuitive interface and thanks to social media, people are already familiar with the concept. Users upload and tag images of the people they want their camera to recognize on their image gallery and simply tap on the picture to setup customized notification or grant access to doors using our facelock feature.
This simple workflow ensures that the user is always in control of information flow with the classification of data into sections that’s curated in a glanceable way for the user. Given the fact that we are the best at converting live-videos into actionable data, we have also created an API, which helps so other smart devices can plugin to create that ecosystem that is so needed in smart home market.