Canela is a simple convolutional neural network library. It's main goal is to be easy to understand. It is, by design, completely unoptimized. The code is easy to understand but pretty slow.
To setup Canela, first setup your environment
make
Then, to start hacking:
source env.sh
To run the tests and build the examples:
make all
You will need a C++ 11 compiler.
Canela relies on googletest
and uses the build system from libarchlab
. The Makefile
will install both of these locally for you.
The utilities depend on and libpng
and libjpeg
which should be installed by default on most systems.
Here's where you'll find the parts of Cenala:
CNN
-- The core Canela source code. CheckoutCNN/README.md
for details.tests
-- the Canela test suiteexamples
-- example code.util
-- Utility and helper functions (e.g., image loaders)datasets
-- sample data sets.
Canela is based on https://github.com/can1357/simple_cnn.