Classifying ImageNet: using the C++ API

Caffe, at its core, is written in C++. It is possible to use the C++API of Caffe to implement an image classification application similarto the Python code presented in one of the Notebook examples. To lookat a more general-purpose example of the Caffe C++ API, you shouldstudy the source code of the command line tool caffe in tools/caffe.cpp.

Presentation

A simple C++ code is proposed inexamples/cpp_classification/classification.cpp. For the sake ofsimplicity, this example does not support oversampling of a singlesample nor batching of multiple independent samples. This example isnot trying to reach the maximum possible classification throughput ona system, but special care was given to avoid unnecessarypessimization while keeping the code readable.

Compiling

The C++ example is built automatically when compiling Caffe. Tocompile Caffe you should follow the documented instructions. Theclassification example will be built as examples/classification.binin your build directory.

Usage

To use the pre-trained CaffeNet model with the classification example,you need to download it from the “Model Zoo” using the followingscript:

  1. ./scripts/download_model_binary.py models/bvlc_reference_caffenet

The ImageNet labels file (also called the synset file) is alsorequired in order to map a prediction to the name of the class:

  1. ./data/ilsvrc12/get_ilsvrc_aux.sh

Using the files that were downloaded, we can classify the provided catimage (examples/images/cat.jpg) using this command:

  1. ./build/examples/cpp_classification/classification.bin \
  2.   models/bvlc_reference_caffenet/deploy.prototxt \
  3.   models/bvlc_reference_caffenet/bvlc_reference_caffenet.caffemodel \
  4.   data/ilsvrc12/imagenet_mean.binaryproto \
  5.   data/ilsvrc12/synset_words.txt \
  6.   examples/images/cat.jpg

The output should look like this:

  1. ---------- Prediction for examples/images/cat.jpg ----------
  2. 0.3134 - "n02123045 tabby, tabby cat"
  3. 0.2380 - "n02123159 tiger cat"
  4. 0.1235 - "n02124075 Egyptian cat"
  5. 0.1003 - "n02119022 red fox, Vulpes vulpes"
  6. 0.0715 - "n02127052 lynx, catamount"

Improving Performance

To further improve performance, you will need to leverage the GPUmore, here are some guidelines:

  • Move the data on the GPU early and perform all preprocessingoperations there.
  • If you have many images to classify simultaneously, you should usebatching (independent images are classified in a single forward pass).
  • Use multiple classification threads to ensure the GPU is always fullyutilized and not waiting for an I/O blocked CPU thread.