Merge pull request movidius#28 from ashwinvijayakumar/rapid-image-cla…

…ssifier

Rapid image classifier

apps/rapid-image-classifier/AUTHORS.txt

@@ -0,0 +1,2 @@
+Name, email, organization
+Ashwin Vijayakumar, [email protected], Intel Corp

apps/rapid-image-classifier/Makefile

@@ -0,0 +1,49 @@
+.PHONY: all
+all: ilsvrc12 mobilenet googlenet squeezenet alexnet
+
+.PHONY: ilsvrc12
+ilsvrc12:
+	@echo "\making ilsvrc12"
+	(cd ../../data/ilsvrc12; make;)
+
+.PHONY: mobilenet
+mobilenet: 
+	@echo "\nmaking mobilenet"
+	(cd ../../tensorflow/mobilenets; make compile;)
+
+.PHONY: googlenet
+googlenet: 
+	@echo "\nmaking googlenet"
+	(cd ../../caffe/GoogLeNet; make compile;)
+
+.PHONY: squeezenet
+squeezenet:
+	@echo "\nmaking squeezenet"
+	(cd ../../caffe/SqueezeNet; make compile;)
+
+.PHONY: alexnet
+alexnet:
+	@echo "\nmaking alexet"
+	(cd ../../caffe/AlexNet; make compile;)
+
+
+.PHONY: run
+run: all
+	@echo "\nRunning rapid-image-classifier.py";
+	python3 rapid-image-classifier.py
+
+.PHONY: help
+help:
+	@echo "possible make targets: ";
+	@echo "  make help - shows this message";
+	@echo "  make all - makes the following: run";
+	@echo "  make run - runs the stream_infer program";
+	@echo "  make clean - removes all created content and temporary files";
+
+clean: clean
+	@echo "\nmaking clean";
+	(cd ../../data/ilsvrc12/; make clean;);
+	(cd ../../tensorflow/mobilenets; make clean;);
+	(cd ../../caffe/GoogLeNet; make clean;);
+	(cd ../../caffe/SqueezeNet; make clean;);
+	(cd ../../caffe/AlexNet; make clean;);

apps/rapid-image-classifier/rapid-image-classifier.py

@@ -0,0 +1,134 @@
+#!/usr/bin/python3
+
+# ****************************************************************************
+# Copyright(c) 2017 Intel Corporation. 
+# License: MIT See LICENSE file in root directory.
+# ****************************************************************************
+
+# How to sequentially classify multiple images using DNNs on 
+# Intel® Movidius™ Neural Compute Stick (NCS)
+
+import mvnc.mvncapi as mvnc
+import numpy
+import skimage
+from skimage import io, transform 
+import os
+from os import listdir, path
+from os.path import expanduser, isfile, join
+
+# User modifiable input parameters
+NCAPPZOO_PATH           = expanduser( '~/workspace/ncappzoo' )
+GRAPH_PATH              = NCAPPZOO_PATH + '/tensorflow/mobilenets/graph'
+IMAGES_PATH             = NCAPPZOO_PATH + '/data/images'
+LABELS_PATH             = NCAPPZOO_PATH + '/tensorflow/mobilenets/categories.txt'
+IMAGE_MEAN              = numpy.float16( 127.5 )
+IMAGE_STDDEV            = ( 1 / 127.5 )
+IMAGE_DIM               = ( 224, 224 )
+
+# ---- Step 1: Open the enumerated device and get a handle to it -------------
+
+def open_ncs_device():
+
+    # Look for enumerated NCS device(s); quit program if none found.
+    devices = mvnc.EnumerateDevices()
+    if len( devices ) == 0:
+        print( 'No devices found' )
+        quit()
+
+    # Get a handle to the first enumerated device and open it
+    device = mvnc.Device( devices[0] )
+    device.OpenDevice()
+
+    return device
+
+# ---- Step 2: Load a graph file onto the NCS device -------------------------
+
+def load_graph( device ):
+
+    # Read the graph file into a buffer
+    with open( GRAPH_PATH, mode='rb' ) as f:
+        blob = f.read()
+
+    # Load the graph buffer into the NCS
+    graph = device.AllocateGraph( blob )
+
+    return graph
+
+# ---- Step 3: Pre-process the images ----------------------------------------
+
+def pre_process_image():
+
+    # Read all images in the folder
+    imgarray = []
+    print_imgarray = []
+
+    onlyfiles = [ f for f in listdir(IMAGES_PATH) 
+                  if isfile( join( IMAGES_PATH, f ) ) ]
+
+    print( "\n\nPre-processing images..." )
+
+    for file in onlyfiles:
+        fimg = IMAGES_PATH + "/" + file
+#        print( "Opening file ", fimg )
+
+        # Read & resize image [Image size is defined during training]
+        img = skimage.io.imread( fimg )
+        print_imgarray.append( skimage.transform.resize( img, ( 700, 700 ) ) ) 
+        img = skimage.transform.resize( img, IMAGE_DIM, preserve_range=True )
+
+        # Convert RGB to BGR [skimage reads image in RGB, but Caffe uses BGR]
+        img = img[:, :, ::-1]
+
+        # Mean subtraction & scaling [A common technique used to center the data]
+        img = img.astype( numpy.float16 )
+        img = ( img - IMAGE_MEAN ) * IMAGE_STDDEV
+
+        imgarray.append( img )
+
+    return imgarray, print_imgarray
+
+# ---- Step 4: Offload images, read & print inference results ----------------
+
+def infer_image( graph, imgarray, print_imgarray ):
+
+    # Load the labels file 
+    labels = numpy.loadtxt( LABELS_PATH, str, delimiter = '\t' )
+
+    print( "\n---- Predictions ----" )
+
+    for index, img in enumerate( imgarray ):
+        # Load the image as a half-precision floating point array
+        graph.LoadTensor( img , 'user object' )
+
+        # Get the results from NCS
+        output, userobj = graph.GetResult()
+        order = output.argsort()[::-1][:6]
+
+        # Print prediction results on the terminal window
+        print( labels[order[0] + 1] )
+
+        # Display the image on which inference was performed
+#        if 'DISPLAY' in os.environ:
+#            skimage.io.imshow( print_imgarray[index] )
+#            skimage.io.show( )
+
+# ---- Step 5: Unload the graph and close the device -------------------------
+
+def close_ncs_device( device, graph ):
+    graph.DeallocateGraph()
+    device.CloseDevice()
+
+def main():
+    device = open_ncs_device()
+    graph = load_graph( device )
+
+    imgarray, print_imgarray = pre_process_image()
+    infer_image( graph, imgarray, print_imgarray )
+
+    close_ncs_device( device, graph )
+
+if __name__ == '__main__':
+    main()
+
+# ==== End of file ===========================================================
+

apps/rapid-image-classifier/readme.md

@@ -0,0 +1,33 @@
+# rapid-image-classifier: Image classification using DNNs on Intel Movidius Neural Compute Stick (NCS)
+
+This directory contains a python3 example that shows how to sequencially classify multiple images using deep neural networks on the Intel Movidius Neural Compute Stick
+
+## Prerequisites
+
+This code example requires that the following components are available:
+1. Movidius Neural Compute Stick
+2. Movidius Neural Compute SDK
+3. Python3
+
+
+## Running the Example
+To run the example code do the following :
+1. Open a terminal and change directory to the rapid-image-classifier base directory
+2. Type the following command in the terminal: make run 
+
+When the application runs normally and is able to connect to the NCS device, the output will be similar to this:
+
+~~~
+Pre-processing images...
+
+---- Predictions ----
+cheetah, chetah, Acinonyx jubatus
+African elephant, Loxodonta africana
+brain coral
+jellyfish
+English setter
+...
+~~~
+
+
+