archive-for-processing / processing-simpleimagelabeling

Image Labeling for Processing using TensorFlow and the Inception model

Simple Image Labeling

Use a pre-trained Inception model with Processing

see examples/Test/Test.pde

TODO

• build a Pi 2/3 optimized version of the JNI library
• look into using different models (different, more labels?)
• attempt to make the code generic so that it can work with different models
• expose GraphBuilder class?

Notes for compiling the JNI bindings for TensorFlow on ARM Linux

The repository comes with libtensorflow_jni.so for armv6hf, but if this needs to be re-done with a later version of TensorFlow:

sudo dd if=/dev/zero of=/dev/swap bs=1M count=1000
sudo mkswap /swap
sudo chmod 0600 /swap
sudo swapon /swap

sudo apt-get install -y autoconf automake libtool gcc-4.8 g++-4.8 zip
sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-4.8 100
sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-4.8 100

sudo cp bezel /usr/local/bin

wget https://github.com/tensorflow/tensorflow/archive/v1.1.0-rc2.tar.gz
tar vfx v1.1.0-rc2.tar.gz
cd tensorflow-1.1.0-rc2

unset IS_MOBILE_PLATFORM in tensorflow/core/platform/platform.h to fix https://github.com/tensorflow/tensorflow/issues/3469

./configure
passed -Os as compile flag, rest default
alternatively, to optimize for Pi2 and Pi3: -mfpu=neon-vfpv4 -funsafe-math-optimizations -ftree-vectorize -Os
something else that was suggested somewhere: -D__ANDROID_TYPES_SLIM__
bazel build --config opt tensorflow/java:tensorflow tensorflow/java:libtensorflow_jni