Device temperature, battery level

The Mapillary app is demanding for mobile devices. Energy hungry hardware like screen, camera, GPS, CPU are busy at the same time and drain the battery.
This energy generates heat in the device and covers and bike/car mounts and protective covers further make it difficult for this heat to compensate into the air.

I have had terrible experiences with the mobile app and warm weather when the app becomes slow, unresponsive, stalls completely or simply crashes. All not mapillarys fault per se, as the OS and device manufacturers throttle speed of the CPU or push a warning when the temperature of camera and CPU are above a threshold. To the user though it is the fault of mapillary. I myself have reported many bugs, some of them turned out to be simply depending on the weather. That’s also a strain on the developers, as those bugs due to high outside temperatures aren’t easy to reproduce (…especially in Sweden).

Here is an idea how to fix this.
Make picture taking frequency (10 instead of 5 m), resolution (lower), focus (manual infinite focus instead of per-image autofocus), display (dim/off) variable and react to the temperature sensors and battery level dynamically. Show some info: reduced image resolution due to hight temp/low battery.

1 Like

Hi
it mainly depends on the mobile you use, e.g. HTC One M9 was horrible in this, as it uses the 810 CPU which created a lot of heat. The honor 7 mobile is fine, even with 35°C heat in summer on bike it works like a charm with 1 picture/sec timeframe.
For the user it is a tough job to see if mobile gets hot, app stotters, if so, reduce frequency of pictures. Or simply get a different, more capable mobile. Or a 2nd cheaper one just for taking pictures.

echelon

1 Like

I have had a deflected made that redirects one of the vents of my air conditioning up onto my cameras.
This seems to work well.

I was using a wiko P4200AN last August in Madagascar for long distance drive using local transportation.

The device was staying on top of the dashboard, under the sun, without air conditioning. Winter(!) outside temperature was higher than +30°C in the shadow. I used low resolution, 10 or 20 m distance and not less than 2 s between pictures. The device was very hot and throttled several times.
Every time the vehicle was stopping, I stopped acquisition, switched of the device and moved it to the shadow. I would appreciate an option, in distance capture mode, so that the device automatically switch off screen and camera when not moving for a few seconds. And switch back on when moving again.

2 Likes

I agree. Suggesting to get a different phone or to gat a car with air-conditioning for a problem that could be solved by software is not a helpful. I did most of the mapping by foot or bike anyways.

HI!

Heating issue cannot be solved by software!
It is a hardware issue with chips and devices are getting used and prdouce heat with usage.
It can only be solved with different hardware or not used by the software.
So: set a lower framerate, lower brightness on your mobile, it will ue less power, use hardware less painfull and create less heat.
Or use a different hardware which will create less heat under the same conditions.
Mapillary is a tool which uses the devices on a very brutal way: take pictures every second, compress it, save it to encrypted storage, show screen all the time, use GPS. Thats a heavy workload and each of these functions create heat, as power is consumed (also the battery creates heat while it is de-charged).

echelon

Ok, my bad wording. The overheating problem can’t be truly solved by software, but it could be alleviated so that the mapillary app stays useful even in a bit warmer then usual conditions.

I find deflecting the air conditioning vent onto the phone/camera works well.

1 Like

+1 with vooni.

Although overheating is a common issue with nowadays smartphones, it is not forbidden to try to fight again it with software improvements instead of buying a much more expensive car and/or smartphone. My suggestions are:

  • auto switch off in distance base mode when not moving
  • with photo min delay activated, wait for the delay before taking picture instead of tacking the picture and skip it if the delay was too small (and reduce refresh rate during the delay?).

(and of course, use low resolution, distance based capture, min delay and manual focus at infinity).

1 Like

You wrote that before. Again: that’s a workaround if you have a car. I mapped 155k images without owning a car.

Just wanted to mention that OpenStreetCam doesn’t make my phone heat up as much as mapillary. You can even run it in the background or switch the screen off while it keeps running. The app matured fast since I last gave it a spin. And since there are so few contributors it’s easy to get on the leaderboard :smiley:

1 Like

I installed Openstreetcam this morning. So far I really like it.
Mapillary could learn a few things from them.

2 Likes

Especially since it is Open Source, mapillary really should be able to learn from it. OSC does have some other weird quirks and I like the backend functionalities of mapillary a lot. I would be sad having to migrate.

About OSC. Has anyone made good pictures in a tunnel yet ? I’d like to see them.

I think that’s hard for the camera because of the lighting situation. That said, if you use OSC with the ODB dongle to the car, it should follow the path of the street and not stop taking pictures when GPS is lost.

Do you mean it cannot be done with a smartphone ? Garmin VIRB XE (whick is a bad camera) can connect with OBD2. But I think it can only be used to show the sensor values in a movie.

I am only talking about the functionality of the app. Think about the distance traveled info from the ODB2 as an additional information to gps that can help interpolate when gps is bad or absent like in a tunnel. It doesn’t concern me much because i’m mostly on a bicycle, but you brought up the topic tunnel, that’s why I am mentioning it.
https://wiki.openstreetmap.org/wiki/OpenStreetView/OBD2

HEy
Yeah, OSC, they try to match all against streets^^ My track of a reailraod ride was funny to see on streets…

echelon

1 Like

True, it leads to really weird paths. But the topic was battery consumption and cpu load, here mapillary can learn from osc.

2 Likes