A C++ example of running TensorFlow Object Detection model in live mode. Inspired by issue.

Requirements (for use without Bazel):

• TensorFlow .so builds (more, requires Bazel to build)
• Eigen3 headers (more)
• OpenCV (more)

Usage:

1. Specify your own paths for necessary libs in CmakeLists.txt
2. Specify your own paths for frozen_inference_graph.pb and labels_map.pbtxt in main.cpp (lines 44-47)
3. Specify your video source (main.cpp, line 80)
4. Have fun

demo/ dir contains frozen graph & labels map from victordibia/handstracking as an example. Feel free to add your own models and contribute.