aertb

AER-toolbox

This library intends to be a minimal tool for loading events from files with common event-camera file extensions into Python.

See the project on PyPI or do pip3 install aertb

Usage

from aertb.core import FileLoader

datLoader = FileLoader('dat') # 'bin', or 'aedat'
datLoader.load_events('../example_data/dat/cars/obj_004414_td.dat')

Supported extensions:

• .dat: N-Cars / Prophesee Cameras
• .bin: N-MNIST, N-Caltech101
• .aedat: PokerDVS
• .mat: DVS-Barrel

It also make the process of loading and iterating HDF5 files easier.

from aertb.core import HDF5File

dataset_train = HDF5File('TRAIN.h5')
train_iterator = dataset_train.iterator(n_samples_group=10, rand=23)

for sample in tqdm(train_iterator):
    # do something with sample.events, sample.label or sample.name

Example: making a GIF

from aertb.core import HDF5File, make_gif

file = HDF5File('../DVS_Barrel.hdf5')
sample = file.load_events(group='moving', name='11')
make_gif(sample, filename='sample_moving.gif', camera_size=(128, 128), n_frames=480, gtype='std')

The library also includes a command line interface for converting files from a given extension to hdf5, as well as gif making capabilities for easy visualisation of the files.

Opening the CLI

• If the install with pip worked perfectly, you can now type aertb in a terminal window and the CLI will open.

• If you are installing it from Github: download you should download the project from github and follow the following instructions:

  • a) git clone ...
  • b) Create a virual environment, if venv is not installed run pip install virtualenv, then python3 -m venv aertb_env
  • c) Run source aertb_env/bin/activate
  • d) Run the following command: pip install -r requirements.txt
  • e) Open the cli with python3 . or with the __main__.py file

Using the CLI

• Once the CLI is open you get a a similar output on your terminal:
• type help to see supported commands and help <topic> to get more info of the command

Examples:

Creating an HDF5 out of a directory

tohdf5 -f 'example_data/dat' -e 'dat' -o 'mytest.h5'

The recommended directory shape is :

 |--Parent (given as parameter)
      |-- LabelClass1
          |-- SampleName1
          |-- SampleName2
          |-- ....
      |-- LabelClass2
          |-- SampleName1
          |-- SampleName2
          |-- ....
      |-- ...

And we suggest that train and test are kept as separate folders so they translate to two different files

Creating an HDF5 out of a single file

tohdf5 -f 'example_data/bin/one/03263.bin' -o 'mytest2.h5'

Creating a gif out of a given file

makegif -f 'example_data/prophesee_dat/test_23l_td.dat' -o 'myGif.gif' -nfr 240 -g 'std'

Exiting the CLI:

• type quit
• Exit virtual environment: $ deactivate