affectiva-api-app

This provides an easy way to extract emotion predictions using Affectiva's JavaScript SDK

Installation

Clone or download the directory

git clone https://github.com/cosanlab/affectiva-api-app.git

Directions for processing images using affectiva_emotion_detector_photo.html

1. Double click the affectiva_emotion_detector_photo.html in Finder or File Browser which should launch a web browser such as Chrome or Firefox.

2. Wait until you see the dectector reports initialized in the logs, and you will see a Choose Files button appear.

3. Choose a file or files with the button. The images will be processed as soon as you click Open.

4. Results of the detector will be combined into output.json file which will be automatically downloaded once processing is complete.

5. You can edit the affectiva_emotion_detector_photo.html to change the output file name, toggle verbosity of logs, or use different detectors (e.g., small or large face).

Directions for processing video using affectiva_emotion_detector.html

1. Open the affectiva_emotion_detector.html in your favorite text editor (e.g. Atom, Sublime) and replace the filename at var filename = 'data/sample_vid.mp4'; with the name of your video filename. It's easies if you have the videos in the affectiva-api-app directory.

2. Modify parameters secs: (default: 0) Beginning time of your video to start feature extraction. sec_step: (default: .1) Increment of the video to extract features. The default of .1 indicates it would process a frame every 100ms. stop_sec: (default: undefined) If you want to process only a portion of your video set stop_sec to the timestamp you wish to stop. verbose: (default: true) When true, the frame being processed, timestamp, and success will be printed on page.

3. Start a webserver on your computer. If using a MAC, open up your terminal. Navigate to the affectiva-api-app folder then start a webserver using

python -m SimpleHTTPServer 8000

4. Now open your favorite browser (e.g. Chrome, FireFox) and type the following to the address bar

http://localhost:8000/

5. You should now see a list of the files in the affectiva-api-app directory. Click on affectiva_emotion_detector.html and the processing will begin.

6. When processing is finished, the browser will automatically download the results file in json format.

7. Read in files An easy way to read in the json files using the following code. You need to add lines=True parameter or it will throw a trailing lines error.

import pandas as pd
df = pd.read_json('~/Downloads/data_sample_vid.json',lines=True)

ToDo's

• Wrapper function / file grabber that can run the extraction for a list of video files.
• Frame-level extraction (currenly analyzes for each second).
• Asynchronous method? (currently recursive until finishes but async/parallel may speed up)
• Electron App?

Files

affectiva_emotion_detector.html This webpage will extract emotions and expression predictions using the Affectiva JavaScript API.

Required files: css
FileSaver.min.js
jquery-3.1.0.slim.js