Marker tracking performance due to light conditions

[lucasroth]

Do you want to request a feature or report a bug? feature, is not a Bug of ar.js

What is the current behavior?

When scanning the target with my celphone motorola X4, the performance is not good in the case of light conditions are "very direct", thus generating glow on the target surface. With other phones this doesn´t happen or is very soft, today I got a lot of shame when it didn´t work in front of possible client, but with his phone (samsung) it did worked. Also this happens with other smartphones, seems like a hardware issue or camera controllers issue, but as I notice there are some configurations available in ar.js, for example I changed the camera settings to 1290x960 and seems to work better (I can´t go to the place where the bug happend to test again).

If the current behavior is a bug, please provide the steps to reproduce.

Use only a direct light to the target, with low or ver high enviorment light. Also the brightness of the paper used to print the marker does affect. Is very difficult to reproduce a light conditions scenario.

Please mention other relevant information such as the browser version, Operating System and Device Name Browser Chrome, operative System Android 9, motorola X4.

What is the expected behavior?

Moreless the same detection capabilities for all smartphones, is not good the case where "on some works good" and in others can not even recongnize the maker.

If this is a feature request, what is motivation or use case for changing the behavior?

The feature is to have a tool to check the interest points of the marker in real time, son can have feedback to the designers and make changes or tunning before deployment. I know about the issues of light conditions and marker materials (mate paper) and design (for interest points). I used to have a CV program that shows in screen the interest points, and was very usefull setting up markerless trackers (using Dfusion CV module). There might be some controllers about the camera exposure than are not well set, or at least for my phone, can´t know how this camera settings affect the detection of interest points, will be great to have a tool to test the interest points and playing with different light conditions scenarios so can design and print the marker properly.

[kalwalt]

Hi @lucasroth with which markers type did you encountered this issue? Pattern, barcode or NFT?

[lucasroth]

Hi @kalwalt sorry for delay on response, I noticed using custom pattern markers. My marker has well performance in good light conditions, but at least on my cellphone goes very bad when direct light makes the marker "shine" due to the paper and ink properties. I have made some changes to the code, from:

<a-marker-camera >
	  <a-plane id="walls" position="2 1 1" color="white" height="1" width="1">	
		<a-entity id="tex1"; text="value: ESCANEA2 EL MARCADOR; align: top; color: black; font: https://cdn.aframe.io/fonts/Aileron-Semibold.fnt; width: 3;" position="-1 0 -5" visible="false" >
	  </a-plane>
	  
	</a-marker-camera>

to this:

<a-entity camera  look-controls="touchEnabled: true">
		<a-marker preset='custom' markerhandler emitevents = "true"  cursor="fuse: false; rayOrigin: mouse;" id="anchor" material arjs-anchor arjs-hit-testing type='pattern' smooth='true' raycaster="objects: .clickable"  url='https://lucas.linuxcorp.cl/AR1/marker/pattern-basenueva2.patt'>
			<a-entity 
				obj-model="obj: #tree-obj; mtl: #tree-mtl" scale="1 1 1" position = '0 0 0' rotation= '0 0 0' class="clickable" id="model" loader>
			</a-entity>		 
		</a-marker>
		<a-plane id="walls2" position="2 1 3" color="white" height="1" width="1">	
			<a-image src="#ground" id="foto1" position = "-2 -1 -4" visible="false" scale= "0.4 0.4 0.4">
		</a-plane>
	</a-entity>

Now it seems to works better, but still not sure on why it happend, and why now seems not to be happening or at least much less.