image_classification.md

April 2, 2018 ยท View on GitHub

Classification for Image

This project supports multiple devices and multiple CNN models for classification. By default, maximum number of devices will be used in inference.
Please make sure you have already set up environment and installed this project correctly. You can refer to the following links for your interested models then.

1 CNN Models

2 Other Arguments


1 CNN Models

Compile NCS graph.

cd /opt/movidius/ncappzoo/caffe/AlexNet
make

Launch object classification service.

roslaunch movidius_ncs_launch ncs_image.launch cnn_type:=alexnet

Launch the example application on another console. For example,

roslaunch movidius_ncs_launch ncs_image_classification_example.launch demo_mode:=0

Compile NCS graph.

cd /opt/movidius/ncappzoo/caffe/GoogleNet
make

Launch object classification service.

roslaunch movidius_ncs_launch ncs_image.launch cnn_type:=googlenet

Launch the example application on another console. For example,

roslaunch movidius_ncs_launch ncs_image_classification_example.launch demo_mode:=0

Compile NCS graph.

cd /opt/movidius/ncappzoo/caffe/SqueezeNet
make

Launch object classification service.

roslaunch movidius_ncs_launch ncs_image.launch cnn_type:=squeezenet

Launch the example application on another console. For example,

roslaunch movidius_ncs_launch ncs_image_classification_example.launch demo_mode:=0

Compile NCS graph.

cd /opt/movidius/ncappzoo/tensorflow/inception_v1
make

Launch object classification service.

roslaunch movidius_ncs_launch ncs_image.launch cnn_type:=inception_v1

Launch the example application on another console. For example,

roslaunch movidius_ncs_launch ncs_image_classification_example.launch demo_mode:=0

Compile NCS graph.

cd /opt/movidius/ncappzoo/tensorflow/inception_v2
make

Launch object classification service.

roslaunch movidius_ncs_launch ncs_image.launch cnn_type:=inception_v2

Launch the example application on another console. For example,

roslaunch movidius_ncs_launch ncs_image_classification_example.launch demo_mode:=0

Compile NCS graph.

cd /opt/movidius/ncappzoo/tensorflow/inception_v3
make

Launch object classification service.

roslaunch movidius_ncs_launch ncs_image.launch cnn_type:=inception_v3

Launch the example application on another console. For example,

roslaunch movidius_ncs_launch ncs_image_classification_example.launch demo_mode:=0

Compile NCS graph.

cd /opt/movidius/ncappzoo/tensorflow/inception_v4
make

Launch object classification service.

roslaunch movidius_ncs_launch ncs_image.launch cnn_type:=inception_v4

Launch the example application on another console. For example,

roslaunch movidius_ncs_launch ncs_image_classification_example.launch demo_mode:=0

Compile NCS graph.

cd /opt/movidius/ncappzoo/tensorflow/mobilenets
make

Launch object classification service.

roslaunch movidius_ncs_launch ncs_image.launch cnn_type:=mobilenet

Launch the example application on another console. For example,

roslaunch movidius_ncs_launch ncs_image_classification_example.launch demo_mode:=0

2 Other Arguments

ArgumentsDescriptionDefault ValueValid Values
device_indexncs device index00~N-1(N is the maximum number of inserted NCS devices)
log_levelncs log level00:Nothing / 1:Errors / 2:Verbose
cnn_typeindicate different cnn modelsgooglenetalexnet / googlenet / squeezenet / inception_v1 / inception_v2 / inception_v3 / inception_v4 / mobilenet
top_nthe number of results to be shown, only valid for classification30~5
demo_modethe display mode of processed results0:sequentially output all results 1: display random results in infinite loop0, 1
image_base_pathdirectory of images to be inferred"/opt/movidius/ncappzoo/data/images/"