AI Multitask Classifiers: From Objects to Emotions

Created by Son Nguyen in 2024, this repository contains Python scripts for various AI-powered classifiers. These classifiers can be used for object detection, face detection, character recognition, and more. The classifiers are built using popular deep learning frameworks such as OpenCV, TensorFlow, and PyTorch.

This repository contains 9 subdirectories for the 9 classifiers:

Classifier	Subdirectory Name
Vehicle Classification	Vehicle-Classification
Human Face Classification	Human-Face-Classification
Mood Classification	Mood-Classification
Flower Classification	Flowers-Classification
Object Classification	Object-Classification
Character Recognition	Character-Recognition
Animal Classification	Animals-Classification
Speech Recognition	Speech-Recognition
Sentiment Analysis	Sentiment-Analysis

Detailed information about each classifier can be found below.

What's even more interesting is that all these classifiers can use your webcam for live testing, video files, or image files!

Please read this README file carefully to understand how to use each classifier and how to run the main script to choose and run any of the classifiers. Happy classifying! 🚀

Table of Contents

• Before You Begin
• Browser Demo
• Main Script - Entry Point
• Flask Web App
• Vehicle Classification
  • Files Included
  • Getting Started
  • Output
• Human Face Classification
  • Files Included
  • Getting Started
  • Output
• Mood Classification
  • Files Included
  • Getting Started
  • Output
• Flower Classification
  • Files Included
  • Getting Started
  • Output
• Object Classification
  • Files Included
  • Getting Started
  • Output
• Character Classification (OCR)
  • Files Included
  • Getting Started

---

Before You Begin

Before you begin, ensure you have the following installed on your machine (run pip install <requirement_name> for each dependency or pip install -r requirements.txt to install all the required packages):

• Python 3.12 or higher (download from the official Python website)
• OpenCV
• TensorFlow
• PyTorch
• NumPy
• Matplotlib
• Tesseract OCR
• Pytesseract
• SpeechRecognition
• MoviePy
• PyDub
• PyAudio
• Scikit-learn
• Pandas
• NLTK
• tqdm
• Joblib
• YoloV3
• jQuery
• Git LFS (for downloading large model weights files)
• A webcam (if you want to use live testing)
• A microphone (if you want to use speech recognition)
• A video file or image file for testing the classifiers
• A stable internet connection (for downloading model weights and dependencies)
• A working speaker or headphones (for speech recognition)
• Additionally, if you would like to train the sentiment classifier, you will need:
  • A machine with sufficient computational resources
  • The large training data files (e.g. training.1600000.processed.noemoticon.csv) or the small dataset generated from it (small_dataset.csv)
• And if you would like to use the website version of this app, you will also need to install Flask and Flask-SocketIO as well.

It is also recommended to use a virtual environment to use these classifiers. You can create a virtual environment using venv or conda:

python -m venv env
source env/bin/activate

[!IMPORTANT] Note: If you are unable to use Git LFS, you can download the necessary files from my Google Drive:

• yolov3.weights
• training.1600000.processed.noemoticon.csv
• test.csv
• train.csv
• small_dataset.csv

Please feel free to let me know if you encounter any problems with any of the files, or with getting started with the project!

[!CAUTION] Update: There has been a known issue with Git LFS bandwidth, which may interrupt your Git cloning experience. Please use the Google Drive links above if you encounter any issues with Git LFS. I apologize for the inconvenience.

---

Browser Demo

Try it in your browser (no install needed!): https://ai-classifiers-demo.vercel.app/

This live demo showcases three tasks:

• Object Classification/Detection — identify common objects from your webcam or an uploaded image.
• Face Detection — locate faces with bounding boxes.
• Character Recognition (OCR) — extract text from images.

Quick start

1. Open the link above (Chromium-based browser recommended).
2. Choose Start Camera (grant permission) or Upload Image.
3. Pick a task and click Run Detection.

Instructions / Troubleshooting

• Use a Chromium-based browser. This demo uses the Shape Detection API.
• Barcode is commonly available by default. Face and Text may require enabling experimental features depending on your build.
• If something is unavailable, try enabling: chrome://flags/#enable-experimental-web-platform-features, restart the browser, then refresh this page.
• Camera requires a secure origin (https:// or localhost).
• Loading an image by URL requires the image server to allow CORS; otherwise you’ll see a SecurityError.
• Detectors are asynchronous and not guaranteed to be real-time. Prefer manual snapshots or occasional checks.

---

Main Script

If you prefer not to navigate through the subdirectories, you can run the main script main.py to choose and run any of the classifiers. The main script will ask you to choose a classifier from the list of available classifiers. You can then select a classifier and run it.

To run the main script, use the following command:

python main.py

The main script will display a list of available classifiers. Enter the number corresponding to the classifier you want to run. The script will then run the selected classifier.

To stop the script, press Q, ESC, or otherwise close the window.

Alternatively, you can also run the individual scripts in each subdirectory below to run the classifiers directly.

---

Flask Web App

If you would like to use the interactive website version of this app, you can run the Flask web app. The web app allows you to use the classifiers through a web interface. You can choose a classifier and the app will run the selected classifier.

To run the Flask web app, use the following command:

python app.py

The web app will start running on http://127.0.0.1:5000/. Open this URL in your web browser to access the web app. You can then choose a classifier from the list of available classifiers and run it. A pop-up window will display the output of the classifier - so be sure to all