Published May 1, 2019 © CC BY-NC-SA

Barbara: The Talking AI Camera

A retro camera captioning every picture.

IntermediateFull instructions provided20 hours2,348

Things used in this project

Hardware components

Raspberry Pi 3 Model B

Raspberry Pi Camera Module

Powerbank

Headphones

Old Camera

Software apps and online services

Raspberry Pi Raspbian

OpenCV

DeepAI

Hand tools and fabrication machines

Tape, Electrical

scissors

Story

Overview

Below a quick overview of the content.

Introduction and showcase video
Camera
Electronics
Code
Result

Project Video

Introduction & Showcase Video

AI being the latest buzzword, this project is all about the slightly useless. After finding a neat API that captions images and an old camera, the goals is set: A camera that captions what it sees!

Camera

This camera is an AGFA Synchro Box from the 50's. Any camera/device will do, as long as a Raspberry Pi, Powerbank and Pi Camera fit inside.

AGFA Synchro Box

Electronics

The electronics are pretty straightforward, to start off setup your Pi and connect the Pi Camera. To power everything you connect the powerbank, the same way you would with your phone.

When everything is wired up, we start on the fiddly part, placing the camera. In this case I kept the shutter open and made sure it was aligned. I takes some time, but hang in there!

Placing and aligning the camera

All that remains is to plug in in our headphones and make sure the Pi plays our audio via the 3.5mm jack.

Plugin in the headphones

Code

The code is a slightly tricky bit, but to make things easy, here is a high level overview:

Take a picture every second
Check if the picture is mostly black (shutter is closed), if so, delete the picture
If not, send the picture to the AI API, receiving the caption back.
This caption is turned into a MP3 file, playing it via the headphones.

For anyone curious, the full code is attached to this article.

Result

Now we have a camera that doesn't take pictures, it only captions them!Some examples for your enjoyment.

A close up of a person holding a cell phone

A close up of a large window with a bird on it

A view of a UNK of a clock tower

Code

import requests
import calendar
import time
from picamera import PiCamera
from PIL import Image
import cv2
import os
from gtts import gTTS

camera = PiCamera()

def speak(text):

	file = "file.mp3"
	tts = gTTS(text, 'en')
	tts.save(file)
	os.system("mpg123 " + file)

def captionImage(imagePath):

    response = requests.post(
        "https://api.deepai.org/api/neuraltalk",
        files={
            'image':  open(imagePath, 'rb')
        },
        headers={'api-key': 'XXXXXXXXXXXXX'}
    )

    return response.json().get("output")

def takePicture(camera):
	
	imageName ='./images/raw/image_' + str(calendar.timegm(time.gmtime()))+'.jpg'
	rawImg = camera.capture(imageName)
	return imageName

while True:
	
	imgPath = takePicture(camera)
	img = Image.open(imgPath)
  
  #Check the min and max colour valeu
	valeus = img.convert("L").getextrema()
  
  #If the brighest colour is below this, shutter must be closed
	if valeus[1] < 160:
		os.remove(imgPath)
	else:
		speak("click!")
		resultCaption = captionImage(imgPath)
		speak(resultCaption)

	
	time.sleep(0.5)

Barbara: The Talking AI Camera

Things used in this project

Hardware components

Software apps and online services

Hand tools and fabrication machines

Story

Overview

Introduction & Showcase Video

Camera

Electronics

Code

Result

Code

Code

Credits

Comments

Embed the widget on your own site

Barbara: The Talking AI Camera

Barbara: The Talking AI Camera

Things used in this project

Hardware components

Software apps and online services

Hand tools and fabrication machines

Story

Overview

Introduction & Showcase Video

Camera

Electronics

Code

Result

Code

Code

Credits

Comments

Related channels and tags