A Python program designed to emulate The Machine from Person of Interest.
- Install Python 3 (must be Python 3.7 or higher)
- Run
pip install -r requirements.txt - Install AVbin (for voice command mode, not required if you don't activate voice mode)
- Configure the system database following the below instructions
- Run
python machine.pyto start the program
To train it to recognize your face, place photos of your face into "/facebase/1/", following the naming scheme "facebase/{subjectnumber}/{filename}.jpg" (e.g. "facebase/1/haroldpic.jpg").
Next you will need to update the subjects.csv file (see more details in the 'designations' section) to add the user to the system, this will only need to be done manually on the first run, after that you can change designations and add users using commands in the program.
Now run the program
The program stores user info (asset number, name and designation) in the subjects.csv file, which can be edited in any spreadsheet program (or notepad, but the formatting is a little harder to understand). The designations are: ADMIN, ANALOG, USER (my current term for indigo agents), THREAT, and UNKNOWN (my stand in for "irrelevant").
When editing this file you must leave the first row as is (with the headings) and start your subjects from the second row onwards (so subject 1's info would be on row 2)
If you are an ADMIN or ANALOG you can also change a subject's designation using the command set {name} as {designation}. For example if I was to type set finch as admin, the program would find the subject with the name finch and set them as an admin.
Typing the command "info" into the console (you must be an ADMIN or ANALOG to use this command) will bring up an information box next to the identified face on the camera stream, similar to the one in the last season.
Typing the command "status" into the console (you must be an ADMIN or ANALOG to use this command) will bring up a status box containing the program uptime and number of subjects detected.
If you are an ANALOG interface or an ADMIN you can type "voice" into the console, this activates the voice command mode, this essentially allows you to use any other command via speech recognition. To go back to typing simply give the command "voice" again.
The program saves mp3 files of each individual word it says, meaning each only needs to be generated once.
To exit the program simply type (or say, if you are in voice mode) "exit".
-
If you get this error:
"Error: Invalid Camera selection, try a different number"
Try increasing your webcam number until it works.
-
If you encounter any problems not listed here, just open an issue.
Python 3.7+ is required.
| Package/module | pip command |
|---|---|
| pyglet (requires AVbin) | pip install pyglet |
| speech_recognition | pip install SpeechRecognition |
| pocketsphinx | pip install pocketsphinx |
| natural.text | pip install natural |
| pyttsx3 | pip install pyvona |
| num2words | pip install num2words |
| Package/module | pip command |
|---|---|
| Pillow | pip install Pillow |
| numpy | pip install numpy |
| deepface | pip install deepface |
| opencv | pip install opencv_python |
| opencv_contrib | pip install opencv_contrib_python |
