Fork of https://github.com/gudovskiy/yoloNCS

YOLO for Intel/Movidius Neural Compute Stick (NCS) demo in MapR environment Run yolo with NCS stick within VirtualBox and stream detected objects to MapR stream Modified code to add the following features :

• Send detected objects to STREAM:TOPIC
• Save image (JPG) when something is detected to Disk (custom path)
• Append a CSV file in addition to the stream (ie. for DB ingest)

VirtualBox setup

• Ubuntu 16.04 TLS Desktop
• VirtualBox 5.2.6 for Mac
• VB extension pack

VM specs

Ubuntu 64 bits 4096MB RAM 2 x vCPU 1 x 25GB SATA HDD (VDI) Audio disabled Network : 1 x NAT + 1 x Host-only USB ports setup : Ports : USB 3.0 xHCI 1st filter : USB2 Movidius 03e7 (vendor ID 03e7) 2nd filter : USB3 Movidius 040e (vendor ID 040e)

After installing Ubuntu

root@mapr:~# sudo su -
root@mapr:~# apt-get install git
root@mapr:~# mkdir -p ~/workspace
root@mapr:~# cd ~/workspace
root@mapr:~/workspace# git clone https://github.com/movidius/ncsdk.git
root@mapr:~/workspace# git clone https://github.com/kromozome2003/MapR-YoloNCS.git
root@mapr:~/workspace# cd ~/workspace/ncsdk
root@mapr:~/workspace/ncsdk# make install

Verify USB stock is present

root@mapr:~/workspace/MapR-YoloNCS# dmesg | grep -i movidius
root@mapr:~/workspace/MapR-YoloNCS# lsusb | grep -i 03e7
root@mapr:~/workspace/ncsdk# usb-devices | grep -i 03e7

![](/images/troubleshoot-usb.png)

Compilation

From your VM

• Compile .prototxt and corresponding .caffemodel (with the same name) to get NCS graph file.

root@mapr:~/workspace/ncsdk# cd ~/workspace/MapR-YoloNCS
root@mapr:~/workspace/MapR-YoloNCS# mkdir -p weights

From your laptop (not the VM)

• Download Pretrained Caffe Models to ./weights/
• Download YOLO_tiny

root@kromozome2003:~/Downloads# scp ~/Downloads/yolo_tiny.caffemodel root@192.168.56.101:/root/workspace/MapR-YoloNCS/weights/

• Check the file is at the right location

root@mapr:~/workspace/MapR-YoloNCS# ls -l weights/

• Compile the model

root@mapr:~/workspace/MapR-YoloNCS# mvNCCompile prototxt/yolo_tiny_deploy.prototxt -w weights/yolo_tiny.caffemodel -s 12
root@mapr:~/workspace/MapR-YoloNCS# ls -l ~/workspace/MapR-YoloNCS/graph

Single Image Script

Run "yolo_example.py" to process a single image.

python3 py_examples/yolo_example.py images/dog.jpg

Try with your webcam (live detection)

Add your webcam to VirtualBox

• Devices->Webcams->Facetime HD Camera

• Run "object_detection_app.py" to process a videos from your camera.

python3 py_examples/object_detection_app.py

• Modify script arguments if needed.

• Press "q" to exit app.

  

Sending detected objects to Kafka stream on MapR platform

python3 py_examples/object_detection_app.py -g http://<USER>:<PASS>@<IP_ADDR>:8082/topics/ -s <STREAM_PATH> -t <TOPIC>

• Modify script arguments axxording to your MapR environment.

• Press "q" to exit app.