Ok, got hold of my HB1080P camera glasses yesterday, and they work nicely with the after-sight unit - especially good to have a comfortable, head-mounted setup.
Just going to need to play around with power-up order/routines, but, no real worries there.
Jacob
Re: Sight for the Blind for <100$
Hi all,
See the below website for some fantastic material on computer vision. I wonder, if we can use content based image retrieval for Semantic scene reading?
http://www.pyimagesearch.com/
See the below website for some fantastic material on computer vision. I wonder, if we can use content based image retrieval for Semantic scene reading?
http://www.pyimagesearch.com/
-
seeingwithsound
- Posts: 165
- Joined: Sun Aug 28, 2011 6:07 am
- Contact: Website
Re: Sight for the Blind for <100$
Glad to hear that, Jacob! Looking forward to your experiences with the camera glasses setup. A training manual is at http://www.seeingwithsound.com/manual/T ... Manual.htmblindzagm wrote:Ok, got hold of my HB1080P camera glasses yesterday, and they work nicely with the after-sight unit - especially good to have a comfortable, head-mounted setup.
Peter
Re: Sight for the Blind for <100$
Thanks, Peter, for the reminder - been a while since went through the training material, and, a 'refresher' is always a good thing...<smile>
Jacob
Jacob
Re: Sight for the Blind for <100$: Going 3D
Hi,
I am working on a system which can create 3D soundscapes.
The system combines depth-sensing elements and 3D audio (binaural audio) synthesis engine for sonification of the 3D depth maps of the environment around a person.
My goal is to make the system wearable which means that single-board computers like Raspberry Pi are potentially a good fit.
I am working on a single-board computer (SBC) -based prototype and plan to evaluate Pi’s performance for this project as well.
Because this thread joins people interested both in the assistive technology for the visually impaired persons and SBCs, I would be grateful for your thoughts related to the system's concept.
I have a short description of my ideas here: https://www.linkedin.com/pulse/assistiv ... r-misharin (the second simulation is the most relevant to the 3D soundscape demonstration).
Alex
I am working on a system which can create 3D soundscapes.
The system combines depth-sensing elements and 3D audio (binaural audio) synthesis engine for sonification of the 3D depth maps of the environment around a person.
My goal is to make the system wearable which means that single-board computers like Raspberry Pi are potentially a good fit.
I am working on a single-board computer (SBC) -based prototype and plan to evaluate Pi’s performance for this project as well.
Because this thread joins people interested both in the assistive technology for the visually impaired persons and SBCs, I would be grateful for your thoughts related to the system's concept.
I have a short description of my ideas here: https://www.linkedin.com/pulse/assistiv ... r-misharin (the second simulation is the most relevant to the 3D soundscape demonstration).
Alex
Re: Sight for the Blind for <100$
Hi Alex,
I would direct you to this link as an example of 3D depth mapping with a visual to sound conversion.
https://www.seeingwithsound.com/binocular.htm
I reviewed your page, and I think your method is interesting because of the high speed. I wonder if you will be able to achieve that on SBC's? what were the specs of the system you used? Have you tried it with actual dual camera input? What kind of device are you using to generate depth information?
As the efforts here are all open source, just adhere to the licence requirements, and you can take what you like.
If you came up with something great, and wanted to share, we would love to add it to the growing stack.
With our goal being creating the most economical device, I do foresee one problem, which is this: generating the depth information generally requires two cameras. Cameras are an expensive component, and adding a camera sounds like it would add a lot of cost.
The impression I've gotten from the speed of the raspberry pi's is that as technology improves, it will make efforts like yours achievable. In fact the speed boost going from the rpi 2 to the rpi 3 has completely bridged the gap for the system we've made. any lag seen on the earlier hardware was pretty much eliminated with the exception of some startup lag.
Unfortunately, it has utterly destroyed battery life. The tradeoffs of power/performance are never fun.
At this time, I would actually prefer to produce with the rpi 2 until things mature more with the rpi 3.
We have certainly made the raspberry pi a wearable device, and would be happy to share.
For more detailed information I would refer you to the git repo which has everything from the software to the hardware designs including custom electronics and 3D printable case designs.
https://github.com/aftersight/After-Sight-Model-1
As a final note, I have already looked into 3D depth map generating devices that are cheap, and have found that it boils down to used kinect sensors. the old version uses structured light, and the new version uses time of flight. Both were found to have outlandish power requirements (ie. the user would need to wear literally pounds of batteries to get useful lifetimes).
I have seen some compact low power consumption devices that are starting to come out, but it appears the price tag is >$500 USD for the sensor without any other components, and the power draw was still high (about 500 mA at 5V).
If cheap depth could generating devices become ubiquitous, my interest level will elevate substantially.
Or if people can find economical dual camera glasses/ultracompact 3D cameras.
Those are my thoughts, I like what you are doing, and if you can achieve the ultrafast refresh rates you demonstrate I think you have a winning idea.
I would direct you to this link as an example of 3D depth mapping with a visual to sound conversion.
https://www.seeingwithsound.com/binocular.htm
I reviewed your page, and I think your method is interesting because of the high speed. I wonder if you will be able to achieve that on SBC's? what were the specs of the system you used? Have you tried it with actual dual camera input? What kind of device are you using to generate depth information?
As the efforts here are all open source, just adhere to the licence requirements, and you can take what you like.
If you came up with something great, and wanted to share, we would love to add it to the growing stack.
With our goal being creating the most economical device, I do foresee one problem, which is this: generating the depth information generally requires two cameras. Cameras are an expensive component, and adding a camera sounds like it would add a lot of cost.
The impression I've gotten from the speed of the raspberry pi's is that as technology improves, it will make efforts like yours achievable. In fact the speed boost going from the rpi 2 to the rpi 3 has completely bridged the gap for the system we've made. any lag seen on the earlier hardware was pretty much eliminated with the exception of some startup lag.
Unfortunately, it has utterly destroyed battery life. The tradeoffs of power/performance are never fun.
At this time, I would actually prefer to produce with the rpi 2 until things mature more with the rpi 3.
We have certainly made the raspberry pi a wearable device, and would be happy to share.
For more detailed information I would refer you to the git repo which has everything from the software to the hardware designs including custom electronics and 3D printable case designs.
https://github.com/aftersight/After-Sight-Model-1
As a final note, I have already looked into 3D depth map generating devices that are cheap, and have found that it boils down to used kinect sensors. the old version uses structured light, and the new version uses time of flight. Both were found to have outlandish power requirements (ie. the user would need to wear literally pounds of batteries to get useful lifetimes).
I have seen some compact low power consumption devices that are starting to come out, but it appears the price tag is >$500 USD for the sensor without any other components, and the power draw was still high (about 500 mA at 5V).
If cheap depth could generating devices become ubiquitous, my interest level will elevate substantially.
Or if people can find economical dual camera glasses/ultracompact 3D cameras.
Those are my thoughts, I like what you are doing, and if you can achieve the ultrafast refresh rates you demonstrate I think you have a winning idea.
Re: Sight for the Blind for <100$
Pranav,
regarding pyimagesearch, I have indeed already purchased Adrian's book. It was where I adapted the code for facial detection from.
Adrian is a wizard at explaining concepts and is currently slowly covering how the convolutional neural networks are created, and how they work, and best of all; How you can do it yourself.
I have to take a break from all of this til I hear back from the Canadian Government about charitable status (and for other reasons too, mostly financial commitment being something I have had to scale back, dayjob demands, family stuff etc...)
should I be wildly successful with having after sight become a charity and have great fundraising success, I would very likely use other portions of Adrian's material to implement some really great applications.
I can't recommend his work more highly to anyone who wants to get exposure to computer vision in an understandable fashion using the very easy high level language of python. It's really great stuff.
regarding pyimagesearch, I have indeed already purchased Adrian's book. It was where I adapted the code for facial detection from.
Adrian is a wizard at explaining concepts and is currently slowly covering how the convolutional neural networks are created, and how they work, and best of all; How you can do it yourself.
I have to take a break from all of this til I hear back from the Canadian Government about charitable status (and for other reasons too, mostly financial commitment being something I have had to scale back, dayjob demands, family stuff etc...)
should I be wildly successful with having after sight become a charity and have great fundraising success, I would very likely use other portions of Adrian's material to implement some really great applications.
I can't recommend his work more highly to anyone who wants to get exposure to computer vision in an understandable fashion using the very easy high level language of python. It's really great stuff.
-
seeingwithsound
- Posts: 165
- Joined: Sun Aug 28, 2011 6:07 am
- Contact: Website
Re: Sight for the Blind for <100$: Going 3D
Hi Alex,almi wrote:Hi,
I am working on a system which can create 3D soundscapes.
Mike has already refered you to the http://www.seeingwithsound.com/binocular.htm page for The vOICe, which also mentions use of the Microsoft Kinect in combination with active window client sonification as used at the University of Sussex to sound live depth maps. You might also be interested in what people in Russia are doing with the After-Sight device,
https://vk.com/voicevision
and maybe add to that in some way or form. Note that sonification of nearby objects with a depth map goes at the expense of detecting and tracking more distant landmarks.
Regards,
Peter
Seeing with Sound - The vOICe
http://www.seeingwithsound.com
Re: Sight for the Blind for <100$
Hi,
Here are some quite general remarks about the depth-sensing systems and 3D audio:
For me it is not a question "if” small, cheap, and (hopefully) power-efficient depth-sensing elements will become available but rather "when" this will happen. And I think this might happen sooner rather then later. You can take a look at the recent MIT device (http://spectrum.ieee.org/tech-talk/semi ... -on-a-chip) or Intel's RealSense cameras which are already included with certain models of laptops. Intel's smartphone kit (https://software.intel.com/en-us/blogs/ ... eloper-kit) has a depth-sensing camera. Google’s project Tango is another recent example. A smartphone or a tablet equipped with a depth-sensing element is a potential option if we are talking about consumer electronics devices as the basis for a system for the blind which has both depth sensing and audio synthesis capabilities.
I think that augmented reality (AR) applications will contribute substantially to driving development of both depth-sensing and 3D audio technology for consumer electronics, including phones and tablets. One can take a look at the Microsoft HoloLens as an example of an AR system that uses both these technologies.
The depth data in my LinkedIn post was produced by Kinect for Xbox One. The audio synthesis engine was running on an old MacBook. The focus of the LinkedIn post is on describing the general ideas and concepts and not on the technical aspects of the technology or its particular implementations. I have added the references to sonification of the depth maps by “The vOICe" to my post.
Thank you for your comments!
Here are some quite general remarks about the depth-sensing systems and 3D audio:
For me it is not a question "if” small, cheap, and (hopefully) power-efficient depth-sensing elements will become available but rather "when" this will happen. And I think this might happen sooner rather then later. You can take a look at the recent MIT device (http://spectrum.ieee.org/tech-talk/semi ... -on-a-chip) or Intel's RealSense cameras which are already included with certain models of laptops. Intel's smartphone kit (https://software.intel.com/en-us/blogs/ ... eloper-kit) has a depth-sensing camera. Google’s project Tango is another recent example. A smartphone or a tablet equipped with a depth-sensing element is a potential option if we are talking about consumer electronics devices as the basis for a system for the blind which has both depth sensing and audio synthesis capabilities.
I think that augmented reality (AR) applications will contribute substantially to driving development of both depth-sensing and 3D audio technology for consumer electronics, including phones and tablets. One can take a look at the Microsoft HoloLens as an example of an AR system that uses both these technologies.
The depth data in my LinkedIn post was produced by Kinect for Xbox One. The audio synthesis engine was running on an old MacBook. The focus of the LinkedIn post is on describing the general ideas and concepts and not on the technical aspects of the technology or its particular implementations. I have added the references to sonification of the depth maps by “The vOICe" to my post.
Thank you for your comments!
Re: Sight for the Blind for <100$
Something have noticed/picked up a few times now is that the after-sight unit will tell me that it has like 5 minutes battery life left, when had it on charge for roundabout 3 hours last night. Anyway, when then switch it off, and on again, it stops warning me about battery life, and carries on working?
Could this just be an OS issue, or could it maybe be a form of connection/contact point issue, and, if so, would it be workable/feasible for me to get a sighted connection to try open up the housing, to just double-check contact points inside, or is this a bad idea?
TIA
Jacob
Could this just be an OS issue, or could it maybe be a form of connection/contact point issue, and, if so, would it be workable/feasible for me to get a sighted connection to try open up the housing, to just double-check contact points inside, or is this a bad idea?
TIA
Jacob
Re: Sight for the Blind for <100$
Ok, along lines of previous reply about possible wiring/contact point issues, took unit through to a friend today to let her try it out, partly since she's also in a wheelchair, and, thus, a hands-free setup would truly suit her the best, and also since included things like distance sensor/obstacle detector would be of special interest to a blind wheelchair user, but, the unit refused to pick up any cameras, no matter how many times I tried powering it off, and on again, plugging in the glasses/camera at different times, etc., so ended up just letting her play around with the radar functionality.
However, when got back home later on, just now tried out testing the unit with a couple of different cameras, making sure it had power from a charging unit, and pretty much the same thing.
Plus, it seemed to repeatedly lock up, and, at times reboot itself, but, finally, when I activated the distance sensor, which was set to vibration output earlier, it locked up into a constant vibration, and just wouldn't do anything, wouldn't react to me turning rotary dial, etc.
Did eventually manage to get it to shut down, by holding in the power button, but, this sort of confirms my prior suspicion that there seem to be issues with contact points/wiring inside the unit.
Either way, I have cloned the git repository, but, think am going to need to get hold of a local electronics person to open it up and have a look for me, and see if there's anything obvious, like a loose connection point, something that needs soldering to be redone, etc. etc.
Also means that, in the meantime, I will switch over to just working with the OS/software installed on a stock raspberry pi B+ v.2 that already had the other version of raspiVoice installed on - couple of days ago had actually already installed the full version onto it to make sure that a keyboard could be used to control the menu system, etc., so will just now need to get hold of a USB numpad to carry around with the unit, and a decent portable mobile phone charger unit for power.
Anyway, if I should specifically not let anyone try open up the after-sight unit, or if I should ask them to look out for anything specific, please let me know, but, am otherwise, at the very least bringing this up, since my guess would be that something like rough handling during shipment, or something might have ended up causing this, since I have looked after the unit pretty carefully since I got my own hands on it, but anyway - suppose this is a standard issue with some forms of electronics?
Jacob
However, when got back home later on, just now tried out testing the unit with a couple of different cameras, making sure it had power from a charging unit, and pretty much the same thing.
Plus, it seemed to repeatedly lock up, and, at times reboot itself, but, finally, when I activated the distance sensor, which was set to vibration output earlier, it locked up into a constant vibration, and just wouldn't do anything, wouldn't react to me turning rotary dial, etc.
Did eventually manage to get it to shut down, by holding in the power button, but, this sort of confirms my prior suspicion that there seem to be issues with contact points/wiring inside the unit.
Either way, I have cloned the git repository, but, think am going to need to get hold of a local electronics person to open it up and have a look for me, and see if there's anything obvious, like a loose connection point, something that needs soldering to be redone, etc. etc.
Also means that, in the meantime, I will switch over to just working with the OS/software installed on a stock raspberry pi B+ v.2 that already had the other version of raspiVoice installed on - couple of days ago had actually already installed the full version onto it to make sure that a keyboard could be used to control the menu system, etc., so will just now need to get hold of a USB numpad to carry around with the unit, and a decent portable mobile phone charger unit for power.
Anyway, if I should specifically not let anyone try open up the after-sight unit, or if I should ask them to look out for anything specific, please let me know, but, am otherwise, at the very least bringing this up, since my guess would be that something like rough handling during shipment, or something might have ended up causing this, since I have looked after the unit pretty carefully since I got my own hands on it, but anyway - suppose this is a standard issue with some forms of electronics?
Jacob
Re: Sight for the Blind for <100$
Hi Jacob,
I have had something similar happen to my unit. Drain the battery, charge it again and you should be good to go.
I have had something similar happen to my unit. Drain the battery, charge it again and you should be good to go.
Re: Sight for the Blind for <100$
In addition to Pranav's advice, if the behaviour persists, rewrite the SD card image.
Regarding cameras not being recognized, try using the USB ports towards the front of the device. The case design has slightly less plastic near those ports and some plugs will work better there.
Also if the camera has its own battery, as many glasses do, it should be fully charged prior to operation with the after sight.
Also, the software should detect when external power is connected. If you don't hear the notification, perhaps the plug is getting worn out on either your charger or the after sight device
I would try these paths first then consider opening it up to inspect for electrical damage
Regarding cameras not being recognized, try using the USB ports towards the front of the device. The case design has slightly less plastic near those ports and some plugs will work better there.
Also if the camera has its own battery, as many glasses do, it should be fully charged prior to operation with the after sight.
Also, the software should detect when external power is connected. If you don't hear the notification, perhaps the plug is getting worn out on either your charger or the after sight device
I would try these paths first then consider opening it up to inspect for electrical damage
Re: Sight for the Blind for <100$
Ok, I drained the battery by getting it going with a desktop USB webcam, and just leaving it running for roundabout 2 hours, and it then seemed to be totally drained, since wouldn't even try fire up.
Then, charged it for a couple of hours, and fired it up again, but, same behaviour.
It does pick up when external power is connected, and notifies me about that if I plug it in, or unplug it while it's running.
The camera glasses were fully charged, and did also get a sighted guy to double check if they had turned on, or not, and I tried both cameras in all four different ports with no real difference occuring.
In terms of rewriting the SDCard, I now switched the after-sight unit's SDCard over to my stock raspberry pi, and it works fine there, but, one issue was when tried inserting the SDCard from that unit into the after-sight, it slipped into the wrong part of the insertion socket, so think it's now inside the unit's main housing, so think will have to get my one sighted connection to try open up the unit to both check connections, and help me get that SDCard back...<smile>
But, on a sort of side note, with the stock raspberry pi implementation of the after-sight OS/img, I got hold of a 6400MAh ultrafast portable power pack, as well as a USB numpad of the genius brand, which then lets me interact with the menus, and, the only bit of functionality now excluded from this separate setup, besides the turning dial/push button and the on/off switch, and, obviously the battery in the housing is the distance sensor, but, will report back when my friend has had a look.
Jacob
Then, charged it for a couple of hours, and fired it up again, but, same behaviour.
It does pick up when external power is connected, and notifies me about that if I plug it in, or unplug it while it's running.
The camera glasses were fully charged, and did also get a sighted guy to double check if they had turned on, or not, and I tried both cameras in all four different ports with no real difference occuring.
In terms of rewriting the SDCard, I now switched the after-sight unit's SDCard over to my stock raspberry pi, and it works fine there, but, one issue was when tried inserting the SDCard from that unit into the after-sight, it slipped into the wrong part of the insertion socket, so think it's now inside the unit's main housing, so think will have to get my one sighted connection to try open up the unit to both check connections, and help me get that SDCard back...<smile>
But, on a sort of side note, with the stock raspberry pi implementation of the after-sight OS/img, I got hold of a 6400MAh ultrafast portable power pack, as well as a USB numpad of the genius brand, which then lets me interact with the menus, and, the only bit of functionality now excluded from this separate setup, besides the turning dial/push button and the on/off switch, and, obviously the battery in the housing is the distance sensor, but, will report back when my friend has had a look.
Jacob
Re: Sight for the Blind for <100$
I suspect for whatever reason your internal battery is probably wrecked. You should be able to check this by using your 6400 mAh external battery.
If you plug in the external battery the system will run off that source, and if the bad behaviour goes away you have a good idea that this is the case.
In other news... snapchat is making video glasses for $130. Which is steep, but the more manufacturers that make video glasses the more likely the quality will go up and the price will come down.
https://news.slashdot.org/story/16/09/2 ... -130-bucks
If you plug in the external battery the system will run off that source, and if the bad behaviour goes away you have a good idea that this is the case.
In other news... snapchat is making video glasses for $130. Which is steep, but the more manufacturers that make video glasses the more likely the quality will go up and the price will come down.
https://news.slashdot.org/story/16/09/2 ... -130-bucks
Re: Sight for the Blind for <100$
Ok, a friend/connection opened it up for me today, and we found what is either a definite sign of, or possibly the cause of the issues.
The battery has a slight melting mark on it, near one corner, but, sort of on the flat surface of the battery, and another piece of the unit - not sure if it's actually electronic, or just part of housing internals, etc., also seemed to have a slight discolouration on it from being near the slight melt on the battery surface.
In other words, going to have to find a local person who has proper electronics experience, and see if they can test parts of the circuitry for me, based on the stuff I pulled of GitHub, and will definitely not try anything relating to powering it up again, until I have replaced the battery.
Jacob
The battery has a slight melting mark on it, near one corner, but, sort of on the flat surface of the battery, and another piece of the unit - not sure if it's actually electronic, or just part of housing internals, etc., also seemed to have a slight discolouration on it from being near the slight melt on the battery surface.
In other words, going to have to find a local person who has proper electronics experience, and see if they can test parts of the circuitry for me, based on the stuff I pulled of GitHub, and will definitely not try anything relating to powering it up again, until I have replaced the battery.
Jacob
Re: Sight for the Blind for <100$
In the next hardware version I switched to two 18650 batteries in parallel. It should be more than double the life and offer higher reliability.
The 18650 cells are extremely mass produced, and thus we can expect lower failure rates.
Some users already have the new batteries and no complaints yet.
If you do replace the battery, ensure the polarity on the jst connector is the same. The jst connector connects the battery to the power module.
In the meantime you should be able to disconnect the internal battery completely, and try running with your external 6400 mAh battery.
The 18650 cells are extremely mass produced, and thus we can expect lower failure rates.
Some users already have the new batteries and no complaints yet.
If you do replace the battery, ensure the polarity on the jst connector is the same. The jst connector connects the battery to the power module.
In the meantime you should be able to disconnect the internal battery completely, and try running with your external 6400 mAh battery.
Re: Sight for the Blind for <100$
Thanks - ad actually meant to ask about removing battery, and just running off external power.
And, yes, that friend of mine who's also in a wheelchair would specifically love to be able to work with the distance sensor/obstacle detector - while couldn't get rest of it to cooperate this past Thursday, I did manage to get that up and running, and she already loved the concept.
And, I just hope that the over-heating/melting didn't damage anything else, but, will let you know if/when I find out anything else.
Jacob
And, yes, that friend of mine who's also in a wheelchair would specifically love to be able to work with the distance sensor/obstacle detector - while couldn't get rest of it to cooperate this past Thursday, I did manage to get that up and running, and she already loved the concept.
And, I just hope that the over-heating/melting didn't damage anything else, but, will let you know if/when I find out anything else.
Jacob
Re: Sight for the Blind for <100$
Well Gentlemen,
The Canadian Government finally denied After Sight's application for charitable status. We had really been in waiting mode for them to make a decision, and now they have.
After Sight may or may not continue as a corporate entity. The idea was to be able to fundraise more effectively and bring the unit cost down while giving donors a tax writeoff.
In the meantime, although I haven't been writing here much, or developing the device, I have been thinking about it, and the time off from working on it has led me down a thought process that I think is worth talking about:
1. Using the raspberry pi zero v1.3 as a main board
2. Using the raspberry pi camera with a wide angle lens as opposed to the camera glasses.
3. moving the distance sensor to a separate device
4. replacing the rotary encoder with three micro pushbutton switches
5. abandoning integration of an accelerometer/gyrometer
6. abandoning the rpi v3 (for now)
I have done an initial cost estimate using the rpi zero and the total is about $120 USD. Much closer to the $100 goal, and much less than the current ~$220 USD.
The distance sensor will get used with a microcontroller/vibration motor/piezo buzzer on a hand strap. I realize most people in this forum are not overly interested in this device, so we could keep it focused on the rpi zero device.
Sound interesting? Or does it just sound like more headache.
My main reasons for wanting to do this are that there are cost benefits, power consumption is lowered, and the size is reduced.
The Canadian Government finally denied After Sight's application for charitable status. We had really been in waiting mode for them to make a decision, and now they have.
After Sight may or may not continue as a corporate entity. The idea was to be able to fundraise more effectively and bring the unit cost down while giving donors a tax writeoff.
In the meantime, although I haven't been writing here much, or developing the device, I have been thinking about it, and the time off from working on it has led me down a thought process that I think is worth talking about:
1. Using the raspberry pi zero v1.3 as a main board
2. Using the raspberry pi camera with a wide angle lens as opposed to the camera glasses.
3. moving the distance sensor to a separate device
4. replacing the rotary encoder with three micro pushbutton switches
5. abandoning integration of an accelerometer/gyrometer
6. abandoning the rpi v3 (for now)
I have done an initial cost estimate using the rpi zero and the total is about $120 USD. Much closer to the $100 goal, and much less than the current ~$220 USD.
The distance sensor will get used with a microcontroller/vibration motor/piezo buzzer on a hand strap. I realize most people in this forum are not overly interested in this device, so we could keep it focused on the rpi zero device.
Sound interesting? Or does it just sound like more headache.
My main reasons for wanting to do this are that there are cost benefits, power consumption is lowered, and the size is reduced.
-
seeingwithsound
- Posts: 165
- Joined: Sun Aug 28, 2011 6:07 am
- Contact: Website
Re: Sight for the Blind for <100$
Hi Mike,
Well, that must sink in. On the one hand it sucks, but on the other hand I guess it gives you great freedom to make your own decisions without writing lengthy reports for accountability. Welcome back, and absolutely great to hear that you haven't given up. I regularly get questions for a turn-key stand-alone device like yours, and of course I then point these people to After-Sight. I have no clear overview of the technical options and the associated costs. Is the Raspberry Pi zero not too slow for us? What about alternatives like the Pine64? https://www.kickstarter.com/projects/pi ... per-comput Maybe we need a matrix overview with some of the alternatives and their pros and cons in price, performance, power, availability of peripherals, expected market lifetime, etc. Will think some more about the possibilities that you list.
Thanks,
Peter
Seeing with Sound - The vOICe
http://www.seeingwithsound.com
Well, that must sink in. On the one hand it sucks, but on the other hand I guess it gives you great freedom to make your own decisions without writing lengthy reports for accountability. Welcome back, and absolutely great to hear that you haven't given up. I regularly get questions for a turn-key stand-alone device like yours, and of course I then point these people to After-Sight. I have no clear overview of the technical options and the associated costs. Is the Raspberry Pi zero not too slow for us? What about alternatives like the Pine64? https://www.kickstarter.com/projects/pi ... per-comput Maybe we need a matrix overview with some of the alternatives and their pros and cons in price, performance, power, availability of peripherals, expected market lifetime, etc. Will think some more about the possibilities that you list.
Thanks,
Peter
Seeing with Sound - The vOICe
http://www.seeingwithsound.com
-
seeingwithsound
- Posts: 165
- Joined: Sun Aug 28, 2011 6:07 am
- Contact: Website
Re: Sight for the Blind for <100$
It looks like Google is starting to support Raspberry Pi (Raspberry Pi 3 Model B) through their Android Things platform https://developer.android.com/things/ha ... rrypi.html I'm not sure yet what possibilities this will give us and whether it would for instance make it possible to run The vOICe for Android on Raspberry Pi and by extension on future After-Sight devices, but it seems interesting enough to keep a proverbial eye on. If this works it could also offer a way to use UVC compliant camera glasses with (The vOICe for) Android, something that is not currently possible with the vast majority of Android devices.
Peter
The vOICe for Android
http://www.seeingwithsound.com/android.htm
https://play.google.com/store/apps/deta ... OICe.vOICe
Peter
The vOICe for Android
http://www.seeingwithsound.com/android.htm
https://play.google.com/store/apps/deta ... OICe.vOICe
-
seeingwithsound
- Posts: 165
- Joined: Sun Aug 28, 2011 6:07 am
- Contact: Website
Re: Sight for the Blind for <100$
A first experiment shows that The vOICe for Android can indeed run as-is on Raspberry Pi 3 under Android Things, but error messages suggest that camera support is currently still incomplete in the developer preview of Android Things https://plus.google.com/+RonaldVanDerLi ... E4JtJbHWqF and it looks like it could take quite some time before this gets fixed with adequate (backward) compatibility https://code.google.com/p/android/issue ... ?id=230863
Peter
The vOICe for Android
http://www.seeingwithsound.com/android.htm
https://play.google.com/store/apps/deta ... OICe.vOICe
Peter
The vOICe for Android
http://www.seeingwithsound.com/android.htm
https://play.google.com/store/apps/deta ... OICe.vOICe
Re: Sight for the Blind for <100$
I have lent my stock raspberry pi 2 B+ unit to another person, for a week or two, to let them try out the after-sight implementation on it.
Thus, this morning just tried installing the after-sight .img - AfterSightPiV2WithUpdater3Apr2016.img - onto a 32Gb SDCard to try running it on a stock raspberry pi 3 model, but, it just seems to not boot up.
If I have a mini speaker plugged in, I almost think I am hearing something happening in terms of the speaker starting to pick up input - just a sort of power up sound, but, no details, and, after that, nothing else seems to happen in terms of console speech output, etc. etc., and, this was both with a full keyboard plugged in, as well as with or without the camera plugged in, since first wanted to just make sure the image would be compatible with the unit.
Anyone else tried out the above image on a pi 3?
TIA
Jacob
Thus, this morning just tried installing the after-sight .img - AfterSightPiV2WithUpdater3Apr2016.img - onto a 32Gb SDCard to try running it on a stock raspberry pi 3 model, but, it just seems to not boot up.
If I have a mini speaker plugged in, I almost think I am hearing something happening in terms of the speaker starting to pick up input - just a sort of power up sound, but, no details, and, after that, nothing else seems to happen in terms of console speech output, etc. etc., and, this was both with a full keyboard plugged in, as well as with or without the camera plugged in, since first wanted to just make sure the image would be compatible with the unit.
Anyone else tried out the above image on a pi 3?
TIA
Jacob
Re: Sight for the Blind for <100$
As I have replied in a separate PM,
there are two filesystem image files.
one for the rpi 2, and one for the rpi 3.
They are both located here:
http://www.after-sight.com/support/
the image from April 2016 is for the rpi 2
the image from Feb 2017 is for the rpi 3.
Although I have not been doing a lot of work on this lately, other people continue to do so. There is a gentleman who is working on integrating the bluetooth on the rpi 3.
At the current time he has been able to get audio devices to connect, but can't get the audio routed through them. He suspects it's some configuration aspect relating to pulseaudio.
If these issues are overcome, then I will probably take it upon myself to write a audio menu driven set of routines to trigger scripts to perform bluetooth pairing and audio routing. For those with the rpi 3, this will eliminate wires.
I shall keep this forum informed the minute such a thing becomes possible.
there are two filesystem image files.
one for the rpi 2, and one for the rpi 3.
They are both located here:
http://www.after-sight.com/support/
the image from April 2016 is for the rpi 2
the image from Feb 2017 is for the rpi 3.
Although I have not been doing a lot of work on this lately, other people continue to do so. There is a gentleman who is working on integrating the bluetooth on the rpi 3.
At the current time he has been able to get audio devices to connect, but can't get the audio routed through them. He suspects it's some configuration aspect relating to pulseaudio.
If these issues are overcome, then I will probably take it upon myself to write a audio menu driven set of routines to trigger scripts to perform bluetooth pairing and audio routing. For those with the rpi 3, this will eliminate wires.
I shall keep this forum informed the minute such a thing becomes possible.
Re: Sight for the Blind for <100$
Ok, installed that image, and it still seemed silent when booted it up with both keyboard and camera connected, so powered it up with only LAN cable connected to my online router, which meant could then get it's IP address, and log in via PuTTY, just to make sure that it was booting up into OS, and, seemed fine.
So then just connected keyboard, but, not camera glasses and it did then boot up into menu system, and if then connected camera, I could toggle raspivoice, etc. on.
So, have now put camera glasses on charge, in case their battery was flat or something, and will try out a couple of different boot-up connection timings later, but, main thing is it can operate...<smile>
Thanks
Jacob Kruger
So then just connected keyboard, but, not camera glasses and it did then boot up into menu system, and if then connected camera, I could toggle raspivoice, etc. on.
So, have now put camera glasses on charge, in case their battery was flat or something, and will try out a couple of different boot-up connection timings later, but, main thing is it can operate...<smile>
Thanks
Jacob Kruger