SSD Inception V2 Gesture Network

Introduction

The SSD Inception V2 network can be used to detect a number of objects specified by a particular training set. This model in particular can detect the following gestures:

point up
point down
point left
point right
exposed palm
closed fist with exposed bottom palm
fist with exposed knuckles.

The provided Makefile does the following:

Downloads a trained model
Downloads test images
Compiles the network using the OpenVINO Model Optimizer
There is a python example (run.py) which runs an inference for all of the test images to show how to use the network with the OpenVINO toolkit.

This network is based on the TensorFlow Object Detection API SSD Inception V2 model. The model was modified, trained, and saved in order to be compatible with the OpenVINO toolkit.

Running this example

make run

The example runs inference using a video stream taken from an existing camera device. Bounding boxes and their associated probabilities and classifications are rendered on frames of the video stream and presented on a OpenCV frame.

Makefile

Provided Makefile describes various targets that help with the above mentioned tasks.

make run

Runs a sample application with the FP16 network. Users must plug in their Intel Neural Compute Stick 2 in order to successfully run this application.

make run_py

Runs the ssd_inception_v2_gesture.py script which takes in a video stream from an existing camera, runs inference on each frame, and renders the output to the user.

make run_FP32

Runs the ssd_inception_v2_gesture.py script with the FP32 network. Note that this application will only run successfully on CPUs that OpenVINO supports.

make run_FP16

Runs the ssd_inception_v2_gesture.py script with the FP16 network. Users must plug in their Intel Neural Compute Stick 2 in order to successfully run this application.

make train

TO BE IMPLEMENTED. Trains a SSD Inception V2 model using the Tensorflow Object Detection API given an Annotations and JPEGImages folder containing .xml and .jpg images, respectively, for training. Training is not necessary since the sample will download a pre-trained model. This option allows for the user to further refine the SSD Inception V2 model if they so desire.

make help

Shows makefile possible targets and brief descriptions.

make all

Makes the follow items: deps, data, compile_model.

make compile_model

Compiles the trained model to generate a OpenVINO IR file. This file can be loaded on the Neural Compute Stick for inferencing.

make get_model

Downloads the trained model.

make deps

Downloads and prepares a trained network for compilation with the OpenVINO toolkit.

make install-reqs

Checks required packages that aren’t installed as part of the OpenVINO installation.

make uninstall-reqs

Uninstalls requirements that were installed by the sample program.

make clean

Removes all the temporary and target files that are created by the Makefile.

Credits

Attributions

ssd_inception_v2_gesture.py is provided per the ATTRIBUTIONS.md file. The rest of the project is provided via the license in the ncappzoo repository’s base directory.