While I was playing with the DVS128 camera (I haven't got many chances yet, as we're still working with our simulator at the moment), I have noticed that there are some annoying parts of pixels that look saturated all the time if I face the camera to bright objects (e.g., monitor) as you can see in the attached screenshots (red circled ones). I tried to adjust the biases settings with a hope that I could optimise its output fits to our environment, but it was a bit difficult as it seems there are a lot of combinations!
Could you give me some advice regarding this? Thank you for your help in advance.
You're probably seeing the flicker on your monitor backlight and the LEDs on your keyboard, typically at a few hundred Hz.
There's not much to do about it except I can offer the following suggestions
1) reduce the pixel bandwidth by starting from the biases DVS128Slow.xml and they trying to reduce the biases Pr and SfBias. However there is a limit as to how slow you can make them.
2) close down the lens aperture. The pixel bandwidth is proportional to illumination.
3) turn up your laptop screen brightness to maximum so that the PWM brightness control doesn't make the backlight or LEDs flicker.
Please provide feedback on your results. Thanks.
Last edit: Tobi Delbruck 2013-10-21
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Thank you for posting my question here for me, here are few additional comments that I would like to add to clarify my issue.
1) When I saw this saturated pixels, I thought about monitors' refresh rate, however I soon found that it doesn't produce similar pixels on the other monitor which has same setting like 1920x1080 @60Hz. I attached a screenshot for this indicating a blue coloured monitor for without and a red coloured monitor for with saturated pixels.
2) The saturated pixels on the keyboard were not caused by the LEDs on it, I think it was a reflection of the monitor screen. I attached another screenshot that a part of the monitor screen was projected on a shiny cup.
As I just started to play with a DVS128, I will try various things, and will post any result regarding this issue later.
I sincerely appreciate your efforts on this jAER open-source project and commercialising the DVS128 camera, and hope I can do great researches using this.
the problem is not the monitor refresh rate, it is the backlight that is flickering. Most laptops regulate the brightness of the screen by switching on and off the backlight at 100Hz or so. Increasing the brightness setting will increase the duty cycle. Usually, if you increase the brightness to maximal level, the backlight is always on, an the DVS will not be saturated.
Cheers
rapha
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Thank you so much for pointing this out. As you suggested, I increased the brightness to maximum level, and the saturated pixels disappeared. Thank you for your help again.
Hanme.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Usually, changing display luminance in normal LCD monitors is achieved by pulse-width-modulation of the backlight. The PWM frequency is usually in the range of a few hundreds of Hertz. This flicker is easily picked up by the DVS/ATIS when using an LCD monitor as input source often resulting in undesired saturation effects of the sensor.
I've found a very simple way of getting rid of the flickering by setting the PWM duty cycle to 100%. In most displays this can be achieved by setting the brightness to the maximum level.
An easy way to determine the PWM frequency is by taking a photo of your monitor, displaying a thin white line on a black background, while moving the camera. Given the shutter speed, the frequency can be easily determined by counting the lines in the resulting image.
Attached is an example from my BenQ 24" monitor. On the left, the image is shown taken with a shutter speed of 1/25 while the display brightness was set to 50%. You can clearly see 8 lines therefore the PWM frequency equals 25*8=200Hz. After setting the brightness to 100%, the duty cycle is also set to 100% which results in a smooth image (non-flickering) shown on the right.
Before setting the brightness to its maximum, I received a 1MEvent/s stream when directing the DVS towards the display. Now, I receive a 0.1KEvent/s stream which is equal to the background activity of a static scene with my current bias settings.
It seems that some displays use another technique to dim the backlight which is not based on any form of PWM (e.g. Apple's retina displays). If you're the lucky owner of such a display you can directly use it as stimuli source for your sensor without changing anything. Otherwise the method described above might be very useful to you.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
(posted by Tobi Delbruck for new user)
While I was playing with the DVS128 camera (I haven't got many chances yet, as we're still working with our simulator at the moment), I have noticed that there are some annoying parts of pixels that look saturated all the time if I face the camera to bright objects (e.g., monitor) as you can see in the attached screenshots (red circled ones). I tried to adjust the biases settings with a hope that I could optimise its output fits to our environment, but it was a bit difficult as it seems there are a lot of combinations!
Could you give me some advice regarding this? Thank you for your help in advance.
Last edit: Tobi Delbruck 2013-10-21
You're probably seeing the flicker on your monitor backlight and the LEDs on your keyboard, typically at a few hundred Hz.
There's not much to do about it except I can offer the following suggestions
1) reduce the pixel bandwidth by starting from the biases DVS128Slow.xml and they trying to reduce the biases Pr and SfBias. However there is a limit as to how slow you can make them.
2) close down the lens aperture. The pixel bandwidth is proportional to illumination.
3) turn up your laptop screen brightness to maximum so that the PWM brightness control doesn't make the backlight or LEDs flicker.
Please provide feedback on your results. Thanks.
Last edit: Tobi Delbruck 2013-10-21
Thank you for posting my question here for me, here are few additional comments that I would like to add to clarify my issue.
1) When I saw this saturated pixels, I thought about monitors' refresh rate, however I soon found that it doesn't produce similar pixels on the other monitor which has same setting like 1920x1080 @60Hz. I attached a screenshot for this indicating a blue coloured monitor for without and a red coloured monitor for with saturated pixels.
2) The saturated pixels on the keyboard were not caused by the LEDs on it, I think it was a reflection of the monitor screen. I attached another screenshot that a part of the monitor screen was projected on a shiny cup.
As I just started to play with a DVS128, I will try various things, and will post any result regarding this issue later.
I sincerely appreciate your efforts on this jAER open-source project and commercialising the DVS128 camera, and hope I can do great researches using this.
Thanks.
It seems it allows only 1 attached file per a post, so here is another screenshot that I mentioned.
Dear Hanme,
the problem is not the monitor refresh rate, it is the backlight that is flickering. Most laptops regulate the brightness of the screen by switching on and off the backlight at 100Hz or so. Increasing the brightness setting will increase the duty cycle. Usually, if you increase the brightness to maximal level, the backlight is always on, an the DVS will not be saturated.
Cheers
rapha
Dear Rapha,
Thank you so much for pointing this out. As you suggested, I increased the brightness to maximum level, and the saturated pixels disappeared. Thank you for your help again.
Hanme.
From earlier suggestion by a user:
Usually, changing display luminance in normal LCD monitors is achieved by pulse-width-modulation of the backlight. The PWM frequency is usually in the range of a few hundreds of Hertz. This flicker is easily picked up by the DVS/ATIS when using an LCD monitor as input source often resulting in undesired saturation effects of the sensor.
I've found a very simple way of getting rid of the flickering by setting the PWM duty cycle to 100%. In most displays this can be achieved by setting the brightness to the maximum level.
An easy way to determine the PWM frequency is by taking a photo of your monitor, displaying a thin white line on a black background, while moving the camera. Given the shutter speed, the frequency can be easily determined by counting the lines in the resulting image.
Attached is an example from my BenQ 24" monitor. On the left, the image is shown taken with a shutter speed of 1/25 while the display brightness was set to 50%. You can clearly see 8 lines therefore the PWM frequency equals 25*8=200Hz. After setting the brightness to 100%, the duty cycle is also set to 100% which results in a smooth image (non-flickering) shown on the right.
Before setting the brightness to its maximum, I received a 1MEvent/s stream when directing the DVS towards the display. Now, I receive a 0.1KEvent/s stream which is equal to the background activity of a static scene with my current bias settings.
It seems that some displays use another technique to dim the backlight which is not based on any form of PWM (e.g. Apple's retina displays). If you're the lucky owner of such a display you can directly use it as stimuli source for your sensor without changing anything. Otherwise the method described above might be very useful to you.