xref: /pdk/apps/CameraITS

# Copyright 2013 The Android Open Source Project
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#      http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.


Android Camera Imaging Test Suite (ITS)
=======================================

1. Introduction
---------------

The ITS is a framework for running tests on the images produced by an Android
camera. The general goal of each test is to configure the camera in a desired
manner and capture one or more shots, and then examine the shots to see if
they contain the expected image data. Many of the tests will require that the
camera is pointed at a specific target chart or be illuminated at a specific
intensity.

2. Setup
--------

There are two components to the ITS:
1. The Android device running ItsService.apk.
2. A host machine connected to the Android device that runs Python tests.

2.1. Device setup
-----------------

Build and install ItsService.apk for your device. After setting up your
shell for Android builds, from the pdk/apps/CameraITS directory run the
following commands:

    cd service
    mma -j32
    adb install -r <YOUR_OUTPUT_PATH>/ItsService.apk

using whatever path is appropriate to your output ItsService.apk file.

2.2. Host PC setup
------------------

The first pre-requisite is the Android SDK, as adb is used to communicate with
the device.

The test framework is based on Python on the host machine. It requires
Python 2.7 and the scipy/numpy stack, including the Python Imaging Library.

(For Ubuntu users)

    sudo apt-get install python-numpy python-scipy python-matplotlib

(For other users)

All of these pieces can be installed on your host machine separately,
however it is highly recommended to install a bundled distribution of
Python that comes with these modules. Some different bundles are listed
here:

    http://www.scipy.org/install.html

Of these, Anaconda has been verified to work with these scripts, and it is
available on Mac, Linux, and Windows from here:

    http://continuum.io/downloads

Note that the Anaconda python executable's directory must be at the front of
your PATH environment variable, assuming that you are using this Python
distribution. The Anaconda installer may set this up for you automatically.

Once your Python installation is ready, set up the test environment.

2.2.1. Linux + Mac OS X
-----------------------

On Linux or Mac OS X, run the following command (in a terminal) from the
pdk/apps/CameraITS directory, from a bash shell:

    source build/envsetup.sh

This will do some basic sanity checks on your Python installation, and set up
the PYTHONPATH environment variable.

2.2.2. Windows
--------------

On Windows, the bash script won't run (unless you have cygwin (which has not
been tested)), but all you need to do is set your PYTHONPATH environment
variable in your shell to point to the pdk/apps/CameraITS/pymodules directory,
giving an absolute path. Without this, you'll get "import" errors when running
the test scripts.

3. Python framework overview
----------------------------

The Python modules are under the pymodules directory, in the "its" package.

* its.device: encapsulates communication with ItsService.apk service running
  on the device
* its.objects: contains a collection of functions for creating Python objects
  corresponding to the Java objects which ItsService.apk uses
* its.image: contains a collection of functions (built on numpy arrays) for
  processing captured images
* its.error: the exception/error class used in this framework
* its.target: functions to set and measure the exposure level to use for
  manual shots in tests, to ensure that the images are exposed well for the
  target scene
* its.dng: functions to work with DNG metadata

All of these module have associated unit tests; to run the unit tests, execute
the modules (rather than importing them).

3.1. Device control
-------------------

The its.device.ItsSession class encapsulates a session with a connected device
under test (which is running ItsService.apk). The session is over TCP, which is
forwarded over adb.

As an overview, the ItsSession.do_capture() function takes a Python dictionary
object as an argument, converts that object to JSON, and sends it to the
device over tcp which then deserializes from the JSON object representation to
Camera2 Java objects (CaptureRequests) which are used to specify one or more
captures. Once the captures are complete, the resultant images are copied back
to the host machine (over tcp again), along with JSON representations of the
CaptureResult and other objects that describe the shot that was actually taken.

The Python capture request object(s) can contain key/value entries corresponding
to any of the Java CaptureRequest object fields.

The output surface's width, height, and format can also be specified. Currently
supported formats are "jpg", "raw", "raw10", "dng", and "yuv", where "yuv" is
YUV420 fully planar. The default output surface is a full sensor YUV420 frame.

The metadata that is returned along with the captured images is also in JSON
format, serialized from the CaptureRequest and CaptureResult objects that were
passed to the capture listener, as well as the CameraProperties object.

3.2. Image processing and analysis
----------------------------------

The its.image module is a collection of Python functions, built on top of numpy
arrays, for manipulating captured images. Some functions of note include:

    load_yuv420_to_rgb_image
    apply_lut_to_image
    apply_matrix_to_image
    write_image

The scripts in the tests directory make use of these modules.

Note that it's important to do heavy image processing using the efficient numpy
ndarray operations, rather than writing complex loops in standard Python to
process pixels. Refer to online docs and examples of numpy for information on
this.

3.3. Tests
----------

The tests directory contains a number of self-contained test scripts. All
tests should pass if the tree is in a good state.

Most of the tests save various files in the current directory. To have all the
output files put into a separate directory, run the script from that directory,
for example:

    mkdir out
    cd out
    python ../tests/scene1/test_linearity.py

Any test can be specified to reboot the camera prior to capturing any shots, by
adding a "reboot" or "reboot=N" command line argument, where N is the number of
seconds to wait after rebooting the device before sending any commands; the
default is 30 seconds.

    python tests/scene1/test_linearity.py reboot
    python tests/scene1/test_linearity.py reboot=20

It's possible that a test could leave the camera in a bad state, in particular
if there are any bugs in the HAL or the camera framework. Rebooting the device
can be used to get it into a known clean state again.

Each test assumes some sort of target or scene. There are multiple scene<N>
folders under the tests directory, and each contains a README file which
describes the scene for the scripts in that folder.

By default, camera device id=0 is opened when the script connects to the unit,
however this can be specified by adding a "camera=1" or similar argument to
the script command line. On a typical device, camera=0 is the main (rear)
camera, and camera=1 is the front-facing camera.

    python tests/scene1/test_linearity.py camera=1

The tools/run_all_tests.py script should be executed from the top-level
CameraITS directory, and it will run all of the tests in an automated fashion,
saving the generated output files along with the stdout and stderr dumps to
a temporary directory.

    python tools/run_all_tests.py

This can be run with the "noinit" argument, and in general any args provided
to this command line will be passed to each script as it is executed.

The tests/inprog directory contains a mix of unfinished, in-progress, and
incomplete tests. These may or may not be useful in testing a HAL impl.,
and as these tests are copmleted they will be moved into the scene<N> folders.

When running individual tests from the command line (as in the examples here),
each test run will ensure that the ItsService is running on the device and is
ready to accept TCP connections. When using a separate test harness to control
this infrastructure, the "noinit" command line argument can be provided to
skip this step; in this case, the test will just try to open a socket to the
service on the device, and will fail if it's not running and ready.

    python tests/scene1/test_linearity.py noinit

3.4. Target exposure
--------------------

The tools/config.py script is a wrapper for the its.target module, which is
used to set an exposure level based on the scene that the camera is imaging.
The purpose of this is to be able to have tests which use hard-coded manual
exposure controls, while at the same time ensuring that the captured images
are properly exposed for the test (and aren't clamped to white or black).

If no argument is provided, the script will use the camera to measure the
scene to determine the exposure level. An argument can be provided to hard-
code the exposure level.

    python tools/config.py
    python tools/config.py 16531519962

This creates a file named its.target.cfg in the current directory, storing the
target exposure level. Tests that use the its.target module will be reusing
this value, if they are run from the same directory and if they contain the
"target" command line argument:

    python tests/scene1/test_linearity.py target

If the "target" argument isn't present, then the script won't use any cached
its.target.cfg values that may be present in the current directory.

3.5. Docs
---------

The pydoc tool can generate HTML docs for the ITS Python modules, using the
following command (run after PYTHONPATH has been set up as described above):

    pydoc -w its its.device its.image its.error its.objects its.dng its.target

There is a tutorial script in the tests folder (named tutorial.py). It
illustrates a number of the its.image and its.device primitives, and shows
how to work with image data in general using this infrastructure. (Its code
is commented with explanatory remarks.)

    python tests/tutorial.py

3.6. List of command line args
---------------------------------

The above doc sections describe the following command line arguments that may
be provided when running a test:

    reboot
    reboot=N
    target
    noinit
    camera=N

4. Known issues
---------------

The Python test scripts don't work if multiple devices are connected to the
host machine; currently, the its.device module uses a simplistic "adb -d"
approach to communicating with the device, assuming that there is only one
device connected. Fixing this is a TODO.