christian-vorhemus / azure-percept-py Goto Github PK

Add a simulation flag so that the code can also be used (and tested) on servers without an actual Percept.

Hi @christian-vorhemus !

Following up on the discussion in pypa/auditwheel#354 I was wondering whether it would make sense to pool our resources to maintain FFmpeg builds for Python projects.

To be more specific, in order to cut down on the build time for PyAV wheels for Linux / Mac / Windows, I maintain this:

https://github.com/PyAV-Org/pyav-ffmpeg

These are builds of FFmpeg + its dependencies, built via cibuildwheel to ensure they are using the exact toolchains required for linking them in Python wheels. The output is a tarball per platform contain include and lib directories.

Windows is a special case, I have not (yet?) been able to compile FFmpeg and its dependencies from source, so I rely on third party builds.

Any thoughts?

The problem appears when running the sample code from the README. The error varies (running the same code) and may also include malloc(): corrupted unsorted chunks 3 or free(): corrupted unsorted chunks

When I install the device, I always get "Operation time out"

1. I connect the USB
2. I allow udev access by adding SUBSYSTEM=="usb", ATTR{idVendor}=="045e", ATTR{idProduct}=="0673", MODE="666" and reloading it
3. I run the Azure Ear example

Stack:

➜ python azure-percept-test.py
Authenticating sensor...
Exception in thread Thread-2:
Traceback (most recent call last):
  File "/home/user/.pyenv/versions/3.8.12/lib/python3.8/threading.py", line 932, in _bootstrap_inner
    self.run()
  File "/home/user/.pyenv/versions/3.8.12/lib/python3.8/threading.py", line 870, in run
    self._target(*self._args, **self._kwargs)
  File "/home/user/.pyenv/versions/3.8.12/lib/python3.8/site-packages/azure/iot/percept/_device_authentication.py", line 94, in start_authentication
    read_msg = dev.read(self.read_endpoint, 5120)
  File "/home/user/.pyenv/versions/3.8.12/lib/python3.8/site-packages/usb/core.py", line 1029, in read
    ret = fn(
  File "/home/user/.pyenv/versions/3.8.12/lib/python3.8/site-packages/usb/backend/libusb1.py", line 846, in bulk_read
    return self.__read(self.lib.libusb_bulk_transfer,
  File "/home/user/.pyenv/versions/3.8.12/lib/python3.8/site-packages/usb/backend/libusb1.py", line 954, in __read
    _check(retval)
  File "/home/user/.pyenv/versions/3.8.12/lib/python3.8/site-packages/usb/backend/libusb1.py", line 602, in _check
    raise USBTimeoutError(_strerror(ret), ret, _libusb_errno[ret])
usb.core.USBTimeoutError: [Errno 110] Operation timed out

Currently the library only supports nets with one tensor as output. Support for multi-tensor outputs should be added.

It seems like the debug level set on the myriad library is set to INFO which results in a lot of logging messages that are put to standard out. It would be cleaner if only errors or critical messages would be shown (potentially on stderr) or - even better - letting the user set the log level.

There are memory leaks detected using tracemalloc and psutil.Process(os.getpid()). Goal of this issue is to diagnose them further and fix them eventually.

Thanks a lot for creating this module to easily contorl camera and audio hardware !
Will it support control audio led in the future?

Thanks.

The function should return the image array fast, at the moment it takes several seconds.

When I create a .wav file but don't close the device (just call stop_recording) the file created is invalid:

formats: can't open input file `./sample.wav': WAVE: RIFF header not found
RuntimeError: Error loading audio file: failed to open file ./sample.wav

Example Code:

audio = AudioDevice()

print("Authenticating sensor...")
while True:
    if audio.is_ready() is True:
        break
    else:
        time.sleep(1)

print("Authentication successful!")

while True:
    audio.start_recording("./sample.wav")
    time.sleep(1)
    audio.stop_recording() # Closes WAV file

# Don't call this to simulate problem
# audio.close()

Reference:

In general the function is buggy as it doesn't return the correct number of frames.

Hi!

Thanks a lot for creating this, amazing that we can use all the sensors this easily!

Currently I am working on a project, but I would love to learn more on how you enabled the Azure ear audio recording. Could you explain in a bit more detail how the audio recording works in this library? Is there a way to do this easily ourselves or do we need to use this library?