4 Comments

Trying out DeepSpeech on a Raspberry Pi 4

{}
 January 3rd, 2020

Deep Speech is an open speech-to-text engine by Mozilla. Speech synthesis and Speech to text are fun to try out, and I read that it could run on a Raspberry Pi4 with ease on one core, so I decided to give it a try.

The Raspberry Pi version is using Google’s TensorFlow Lite for an implementation of Baidu’s DeepSpeech architecture.

Installing it on a Raspberry 4 Buster distribution was not straightforward. First I read instructions on the Github page and tried to download and install the git version and, but I ran into problems. It was taking ages and I ran into the famous `wheels` problem.

Failed building wheel for scipy

After tweaking and trying a few times, i gave up on the Github version and tried the instructions here, but also that was a bumpy road. But success waits in the end.

Let’s go, how to install DeepSpeech on the RPI4

Create a dev directory:

mkdir dev
cd dev

Create a Python Virtual environment.

python3 -m venv deepspeech-train-venv

Activate the virtual environment

source dev/deepspeech-train-venv/bin/activate

Create the deepspeech directory

mkdir deepspeech
cd deepspeech

Install deepspeech

pip install deepspeech

Download pre-trained English model

curl -LO https://github.com/mozilla/DeepSpeech/releases/download/v0.6.0/deepspeech-0.6.0-models.tar.gz
tar xvf deepspeech-0.6.0-models.tar.gz

Download example audio files

curl -LO https://github.com/mozilla/DeepSpeech/releases/download/v0.6.0/audio-0.6.0.tar.gz
tar xvf audio-0.6.0.tar.gz

Done, run, well , eh, I tried to run the example on the instruction page

deepspeech --model deepspeech-0.6.0-models/output_graph.pbmm --lm deepspeech-0.6.0-models/lm.binary --trie deepspeech-0.6.0-models/trie --audio audio/2830-3980-0043.wav

Errors!?!  I installed a missing dependency:

sudo apt install libatlas3-base

Still errors

ModuleNotFoundError: No module named 'numpy.core._multiarray_umath'

So I check if I had numpy installed

pip install numpy
Looking in indexes: https://pypi.org/simple, https://www.piwheels.org/simple
Requirement already satisfied: numpy in /home/pi/dev/deepspeech-train-venv/lib/python3.7/site-packages (1.15.4)

I decided to update numpy:

pip install --upgrade numpy
Looking in indexes: https://pypi.org/simple, https://www.piwheels.org/simple
Collecting numpy
Using cached https://www.piwheels.org/simple/numpy/numpy-1.18.0-cp37-cp37m-linux_armv7l.whl
tensorboard 2.0.2 has requirement setuptools>=41.0.0, but you'll have setuptools 40.8.0 which is incompatible.
Installing collected packages: numpy
Found existing installation: numpy 1.15.4
Uninstalling numpy-1.15.4:
Successfully uninstalled numpy-1.15.4
Successfully installed numpy-1.18.0

So i decided to update setuptools too:

pip install --upgrade setuptools
Looking in indexes: https://pypi.org/simple, https://www.piwheels.org/simple
Collecting setuptools
Using cached https://files.pythonhosted.org/packages/f9/d3/955738b20d3832dfa3cd3d9b07e29a8162edb480bf988332f5e6e48ca444/setuptools-44.0.0-py2.py3-none-any.whl
Installing collected packages: setuptools
Found existing installation: setuptools 40.8.0
Uninstalling setuptools-40.8.0:
Successfully uninstalled setuptools-40.8.0
Successfully installed setuptools-44.0.0

I tried to run the example on the instruction page again
# Transcribe an audio file

deepspeech --model deepspeech-0.6.0-models/output_graph.pbmm --lm deepspeech-0.6.0-models/lm.binary --trie deepspeech-0.6.0-models/trie --audio audio/2830-3980-0043.wav

Another error

Loading model from file deepspeech-0.6.0-models/output_graph.pbmm
TensorFlow: v1.14.0-21-ge77504a
DeepSpeech: v0.6.0-0-g6d43e21
ERROR: Model provided has model identifier '='+;', should be 'TFL3'

Didn’t work. I needed to change the model to `tflite`

deepspeech --model deepspeech-0.6.0-models/output_graph.tflite --lm deepspeech-0.6.0-models/lm.binary --trie deepspeech-0.6.0-models/trie --audio audio/2830-3980-0043.wav

Success in the end!

Loading model from file deepspeech-0.6.0-models/output_graph.tflite
TensorFlow: v1.14.0-21-ge77504a
DeepSpeech: v0.6.0-0-g6d43e21
INFO: Initialized TensorFlow Lite runtime.
Loaded model in 0.0019s.
Running inference.
why should one hault on the way
Inference took 4.091s for 2.735s audio file.

Then I played the audio-file:

aplay audio/4507-16021-0012.wav

Must say DeepSpeech is much smarter then me, I couldn’t understand it:
why should one hault on the way

BTW good question. No I need another engine to answer that!

Way to go, folks.

Posted in raspberrypi, Webtechnology | 4 Comments

13 Comments

Fixing the iPhone CSS hover problem on iOS

{}
 November 6th, 2019

Mobile phones don’t have a mouse, your greasy fat touchy fingers have to do the job, they kinda act like a giant mouse.

There is major difference: a mouse can click, drag and  hover above the screen, while your fingers can click and swipe, but not hover. Well they can, but nothing is happening then. 🙂

The official terms here are: mouse-events, touch events, and pointer events, for both mouse and touch events. So touch events don’t know about hover, cause that is a mouse event.

So those nice old style CSS-only dropdown menu’s won’t work on your phone, because the depend on :hover.

Hover events work on mobile Android or Firefox browsers, as those browser vendors looked for compatibility, and made a click with your fingers act like hover on elements that are not a link. That is smart thinking, and good care for compatibility. Keep old sites working. Keep sites accessible without forcing Javascript.

But they don’t work on iPhone or mobile Safari, so we need a solution for that.

Plain HTML is and has always been responsive. It’s the CSS that destroyed it, by setting explicit width.  Well there was no max-width, so designers had not much choice. But remove all CSS and magically your HTML will be responsive, except for tables and large images, but at least you can scroll them into sight so everything is still accessible.

Pity the Apple people. An iPhone with mobile Safari doesn’t do :hover, breaking compatibility for old sites, and forcing webdevelopers to use `Javascript`  for these trivial things. Annoying.

There are examples on the internet how to fix that, like adding an onclick attribute to every element you want the :hover rule for, but that’s adding a lot of code, and the elements the don’t hide again when you click somewhere else.

We need better solutions!

How to make :hover  work on Safari iOS on iPhones and iPads.

Here are a few very simple options, that I came up with having a new iPhone 11 around for a couple of days. Tested on the latest iOS 13.
Read the rest of this entry »

Posted in browser, CSS3, HTML5, mobile, Webtechnology | 13 Comments

1 Comment

4 ways to connect your Raspberry Pi 4 to the internet

{}
 June 26th, 2019

The just introduced Raspberry Pi 4 has delivered some much wished features: faster internet/network, a gigabit Ethernet connection, and faster USB, 2 USB 3.0 ports. This will open the door to a Raspberry Pi driven NAS solution that will offer high speed in combination with the new USB3 ports.

This post is about exploring the USB OTG  (On The Go) feature of the Raspberry Pi 4. We will find out if it’s possible to power and connect a RPI4 with just a USB cable.

Ethernet is easiest way to connect internet, the second is wireless, using WiFi. The new RPI4 offers dual-band 802.11ac wireless networking. Not as fast as the gigabit Ethernet internet, but it will offer speeds like 100 Mbps, around a 100Mb Ethernet connection. That’s quite speedy too.

The third way is using Bluetooth. Like connecting your laptop to a Bluetooth hot-spot on your phone. That will be a bit slower. Will try that later.

The fourth way is using a trick we know from the smallest Raspberry Pi, the Raspberry Zero. The Raspberry Pi 4 has a USB-C power connector, but that USB-C port is also an USB OTG connector, so the Raspberry Pi can be used as a USB Gadget: the USB port can switch USB states (MASTER/SLAVE or HOST/DEVICE) and be used like a keyboard or mouse, or USB Ethernet modem.

Actually you can power and connect to a Raspberry Pi 4 with just an USB-cable. I tried it with the new Raspbian Buster and it’s working fine. I used the same Samsung cable I used for my Zero (and phone), added an 1 euro micro-USB to USB-C adapter, and it’s just working fine. It seems my laptop is offering enough power for a headless, and armless Pi.

To enable OTG networking, add this to `/boot/config.txt`:

# enable USB OTG
dtoverlay=dwc2

Then add this to `/boot/cmdline.txt` after rootwait:

modules-load=dwc2,g_ether

Reboot your Raspberry pi and convert the new wired connection to `shared` in the connection manager on your laptop (e.g. Ubuntu 18.04), and connect to your new Raspberry Pi 4.

It will have an IP like: 10.42.0.xxx.

If you have added  a empty file `ssh` to the boot sector of your SD card touch /boot/ssh, the Pi will start with SSH enabled.

So you can connect from your laptop with:

ssh pi@raspberrypi.local

For more details, see these posts Connecting to a Raspberry Pi Zero with just an USB cable (I) or Connecting to a Raspberry Pi Zero with just an USB cable (II)

I wonder if it will be possible to boot the Raspberry Pi 4 entirely over USB, without a SD card, like the RPI Zero. Not at the moment though.

Posted in raspberrypi | 1 Comment

5 Comments

How to check the Signing Certificate on an Android app / apk.

{}
 January 3rd, 2019

If you own a Android Phone, and you want to use Signal instead of Whatsapp or Telegram for privacy matters, and, for the same privacy matters, you don’t have a Google Account on your phone, or you don’t use Google Play but the free F-Droid, there is a solution. You can download the Signal APK from their website: https://signal.org/android/apk/

They give a warning:

Advanced users with special needs can download the Signal APK directly. Most users should not do this under normal circumstances.

What are normal circumstances these days? You can’t trust Facebook anymore, maybe you can trust Google, but you don’t wanna trust Google, because trusting Google is telling Google where you are, what you do, what you say, what and who you see and who your friends are.

It’s not a matter of trust, it’s a matter of privacy and decency that you don’t do that.

So downloading the Signal APK is probably what you should do these days. But how to be sure you download the real one?

The websites tells you to verify the signing certificate on the APK matches this SHA256 fingerprint. Unfortunately they don’t tell you how to do that.

Verify the signing certificate on the Signal APK.

This one-liner will show you the SHA256 Fingerprint that has to be checked:

unzip -p Signal-website-release-4.31.6.apk META-INF/SIGNAL_S.RSA > /tmp/tmp.cert ; keytool -printcert -file /tmp/tmp.cert

You get this output:

Owner: CN=Whisper Systems, OU=Research and Development, O=Whisper Systems, L=Pittsburgh, ST=PA, C=US
Issuer: CN=Whisper Systems, OU=Research and Development, O=Whisper Systems, L=Pittsburgh, ST=PA, C=US
Serial number: 4bfbebba
Valid from: Tue May 25 17:24:42 CEST 2010 until: Tue May 16 17:24:42 CEST 2045
Certificate fingerprints:
SHA1: 45:98:9D:C9:AD:87:28:C2:AA:9A:82:FA:55:50:3E:34:A8:87:93:74
SHA256: 29:F3:4E:5F:27:F2:11:B4:24:BC:5B:F9:D6:71:62:C0:EA:FB:A2:DA:35:AF:35:C1:64:16:FC:44:62:76:BA:26
Signature algorithm name: SHA1withRSA
Subject Public Key Algorithm: RSA (1024)
Version: 3

As you can see the SHA256 is the same fingerprint as on the Signal download page.

It’s verified. The Signal apk is safe to use now.

UPDATE 2022 (see new blogpost)

Posted in mobile, Webtechnology | 5 Comments

No Comments

Keeping the old TinyMCE editor while upgrading to WordPress 5.0

{}
 December 30th, 2018

WordPress 5 has introduced a new block editor. Playing with the layout is easier this way, but on the other hand, writing a new post should not take more time than typing the text. Layout should be in the theme.

I guess competition from other website-hosting companies like wix.com, yola.com, weebly.com and jimdo.com is the culprit here. They all offer drop and drag editors for layout. No need for any knowledge of HTMl and CSS, or PHP. A noob should do the job apparently. And now WordPress is following. Designing without a developer is the new thing.

Not sure if I like that, WordPress has better performance with a minimum of plugins, and is CSS not handled better in CSS files?

If you’re used to the old, and why not, it’s working fine, just add this to your function file:

add_filter('use_block_editor_for_post_type', 'atys_disable_block_editor');
function atys_disable_block_editor($use_block_editor) {
return false;
}

The new block editor is another example that progress is not necessarily an improvement for all. Some will be scared by all the changes, speaking for myself: I think I will switch to a markdown editor sooner or later.

Less is more. For now, the old editor is OK.

Posted in browser | No Comments

No Comments

Fixing the network after upgrading to Ubuntu 18.04 LTS from 16.04 LTS

December 18th, 2018

Just updated 16.04LTS to 18.04LTS, went easier and smoother than expected.

Under the hood there are major changes in Ubuntu, I decided to stick to Unity rather then the new Gnome 3, it’s much smoother on older hardware. You only change the setting at login once and you done.

The only nasty thing: DNS (Pi-Hole) was not working. I could fix that by manually setting the DNS server in /etc/resolv.conf from `nameserver 127.0.0.53` to the DNS of my router, but that would not persist a restart.

To get info about your network and DNS:


systemd-resolve --status

Link 1 (eth0)
      Current Scopes: none
       LLMNR setting: yes
MulticastDNS setting: no

Current scopes is none, but it should be DNS

Seems there was something wrong with the NetworkManager. Somehow in the past I’ve manually edited `/etc/NetworkManager/NetworkManager.conf` file. Can’t remember when I did that, could be years ago, Ubuntu 10.04 or something ;).

I had to comment out the line that explicitly set `no-auto-default`, meaning don’t let the NetworkManager create a default wired connection for my ethernet card.  

[main]
...
#no-auto-default=<mac-adress>
...
[ifupdown]
managed=false

After commenting it out, NetworkManager could create a default connection and after restart (the daemon), everything was OK. DNS was working.

systemd-resolve --status
Link 1 (eth0)
Current Scopes: DNS
LLMNR setting: yes
MulticastDNS setting: no

As you can see DNS is now working

Maybe this can help you, if you experience the same troubles.

Posted in Webtechnology | No Comments

Older Entries
Newer Entries