In Use (continued):

I set up a Canary in my kid's playroom in the basement for a while too (please excuse the mess). This area gets a lot of activity, as well as a mix of artificial light, low light ( as the sun comes through the small windows), and darkness. This was going to serve as my main test of the IR night vision, which overall worked very well. The image below shows a shot from the mid afternoon where some sun is coming in, but because the room lights are off it had to go to night vision mode.



The next shot is from the same time, but with the camera in landscape mode, instead of portrait mode. What we can see is that the image is not as sharp as the shots taken from the well lit diming room on the previous page. Things in the center of the field of view are fairly clear, but as you get to the edges the quality decreases and you pick up some pixilation and what looks like artifacts within a video game. The lighting is definitely adequate to light up a room in low light or no light, but making a clear identification of someone might require that they enter the middle of the field of view.



While on this camera / view, let's look at the timeline... The below left image shows that activity was detected at 6:50AM, which was just the light coming on for the first time that day. The below right image is merely a landscape view of the same scene. Also worth noting is to confirm that with the lights in the room turned on, that the pixilation and artifacting scene in night vision mode disappear.



I can understand that going from light to dark or dark to light would get flagged as activity initially, but I had a hard time traingin this sort of thing out of the alerts I would receive. The below left image shows the screen where you can try to help the Canary learn what activity is worth alerting you about. The on / off of the lights may be too extreme since it impacts the whole scene, but I also had trouble teaching it to ignore much smaller things. For example, the light coming through the basement window would regularly cast a rectangle of light on the ground about 2 feet by three feet... Which is just a small portion of the entire field of view. When the sun would go behind the clouds, the light spot would fade away, and then return again. These would be tagged and I would be alerted. I tagged them as "sunlight", but the alerts persisted and started to get a bit annoying.



The above right image shows one of my children sneaking through the frame, but not quick enough to avoid detection by the Canary. Again, I can understand that a variety of different sized people with different clothes each day could be hard to train out, but there were plenty of alerts, and the people based alerts weren't the most bothersome.

A finally note on the testing of the cameras, which happened to be during night vision mode, is also a disappointment for me. With the cameras on and the lights off as I was monitoring the scene with the IR lights providing the illumination, I entered the room with the main goal of seeing how noticeable the red from the IR LEDs really was. Compared to many other cameras, the red glow was rather subtle, and while it might be noticeable, it was not as pronounced as the red seen on some of my other cameras (namely Foscam branded units). The problem with this test was that even though I was standing right in front of the Canary, when I looked at the screen of the phone, the room still showed as empty. The screen confirmed that this was a live image, so I waited a bit to see if there was some sort of delay. After about 15 seconds I was getting annoyed, so I turned the lights on in the room... But the Canary still showed a darkened room with no one in it. Alerting me to trivial activity is one thing, but to complete miss activity is not acceptable.