Skip to content

Latest commit

 

History

History
60 lines (47 loc) · 2.55 KB

README.md

File metadata and controls

60 lines (47 loc) · 2.55 KB

SmartReception

This is a part of implement of the smart office. Now, the smart reception is just an single sense which only support to unlock the door if known employees come.

Precondition

  1. an IPC installed in front of the office entrance;
  2. a smart locker should be installed;
  3. both the IPC and the locker and the device which run this code should in the same network;

About the code

In this code, we simply finished the following sense:

  1. real time face detection: two ways are supported which are opencv and dlib;
  2. face recogtition: After detection, the face will be sent to a pre-trained VGG model to estimate who are they with the acceration by Movidius ncs, then, record the results to redis database;
  3. faceSensor: subscribe the recognition result, and envelop the result as a self-defined OCF face sensor. As soon as face recognized, it will notify all subscribers;
  4. faceApp: Observe the faceSensor resource and the smart locker resource as well, if the sensor report the known face, it will post a unlock message to the smart locker to unlock the door, and if not, nothing will be done.

Note: this sense has been performed on RaspberryPI + Intel Movidius ncs + Wulian smart home devices. The smart locker OCF device is not included in this code.

Installation：

  1. install the dlib
  2. install opencv3, python-opencv and associated
  3. install dependency: like numpy, redis, redis-server
  4. install ncs sdk
  5. install for CoAP Server:
    a. pip3 install LinkHeader
    b. install aiocoap
  6. install the iotivity sdk
  7. compile the faceApp and faceSensor

Configuration

url: the rtsp stream url of the IPC
graph: the graph file loadable by Movidius ncs
mean: mean file of the all trained faces;
label: label file of the pre-trained model, here is the names of the faces;
fps: skip how many frames when processing;
resize: for acceration in some low performance platform like raspberry pi, zoom out the size of the frame to be processed;
debug: int value. set to none zero if you want to enable debug;

Run

  1. python3 detect.py -c conf.json
  2. python3 server.py
  3. run the face app and face sensor

Donation

If you like this code or idea, please consider donating to the address below.

fengping [email protected]
1HDHUQGjU42ytfJDud1vGuJb4HksUcATjh

TODO

replace the detection and recognition with the face-comparation or realtime-face-recognition.
connect to the attendance system, which should also detect the off work;
support the legacy unlock device;