Skip to content

Latest commit

 

History

History
636 lines (463 loc) · 25.6 KB

README.md

File metadata and controls

636 lines (463 loc) · 25.6 KB
Raw

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 8 subdirectories: one for vehicle classification, one for human face classification, one for mood classification, one for flower classification , one for object classification, one for character classification, one for animal classification, and one for speech recognition, namely Vehicle-Classification, Human-Face-Classification, Mood-Classification, Flowers-Classification, Object-Classification, Character-Recognition, Animals-Classification, and Speech-Recognition. Refer to the information below for details on each classifier.

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

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
  • 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 file (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.

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

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 allow pop-ups in your browser.

Here is what it looks like:

Flask Web App

Note that the app has also been deployed to Heroku at this link. However, due to changes in Heroku's free tier regarding available Dynos (and I'm a broke college student), the app may not work as expected. If you encounter any issues, please run the app locally using the instructions above.

Vehicle Classification

Files Included

  • coco.names: Class names used for vehicle detection.
  • traffic.mp4: Sample video for vehicle detection.
  • india.jpg: Sample image for vehicle detection.
  • yolov3.cfg: YOLOv3 model configuration file.
  • yolov3.weights: Pre-trained YOLOv3 model weights.
  • vehicle_detection.py: Python script for vehicle detection and classification.

Getting Started

  1. Clone the Repository

    git clone https://github.com/hoangsonww/AI-Classification.git
cd AI-Classification/Vehicle-Classification

  2. Download Model Weights Download the pre-trained YOLOv3 model weights (yolov3.weights) from the official YOLO website or another trusted source and place it in the Vehicle-Classification directory.

  3. Install Dependencies Install the required Python dependencies.

    pip install -r requirements.txt

  4. Install and Pull Git LFS Install Git LFS by following the instructions on the official Git LFS website. Then, pull the model weights using Git LFS.

    git lfs install
git lfs pull

    Alternatively, you can download the weights file from the official YOLO website and place it in the Vehicle-Classification directory. However, using Git LFS is recommended.

    This is crucial for pulling the large model weights file. Without Git LFS, the weights file may not be downloaded correctly and the script may not work as expected.

  5. Run Vehicle Detection

    python vehicle_detection.py

You will then be asked to choose your input type (image, video, or webcam). Enter image to classify the vehicles in the sample video provided (traffic.mp4), or enter video to classify vehicles in a video file. You can also use your webcam for live testing.

All our classifiers will only stop when you press Q, ESC, or otherwise close the window.

Output

The output video will display the detected vehicles along with their class labels. The class labels are based on the COCO dataset, which includes various classes such as car, truck, bus, motorcycle, and bicycle.

Example output:

Vehicle Classification Output

Face Classification

Files Included

  • deploy.prototxt: Model configuration file for the face detector.
  • res10_300x300_ssd_iter_140000.caffemodel: Pre-trained model weights for face detection.
  • age_deploy.prototxt: Model configuration file for age prediction.
  • age_net.caffemodel: Pre-trained model weights for age prediction.
  • gender_deploy.prototxt: Model configuration file for gender prediction.
  • gender_net.caffemodel: Pre-trained model weights for gender prediction.
  • faces_classification.py: Python script for face detection, age, and gender classification.
  • woman-30.mp4: Sample video for face classification
  • man.jpg: Sample image for face classification.

Getting Started

  1. Clone the Repository

    git clone https://github.com/hoangsonww/AI-Classification.git
cd AI-Classification/Face-Classification

  2. Download Model Weights: Ensure you have the model weights (res10_300x300_ssd_iter_140000.caffemodel, age_net.caffemodel, gender_net.caffemodel) in the Human-Face-Classification directory.

  3. Install Dependencies: Install the required Python dependencies.

    pip install -r requirements.txt

  4. Run Face Classification: Execute the face_classification.py script.

    python face_classification.py

You will then be asked to choose your input type (image, video, or webcam). Enter image to classify the faces in the sample image provided (woman-30.mp4), or enter video to classify faces in a video file. You can also use your webcam for live testing.

All our classifiers will only stop when you press Q, ESC, or otherwise close the window.

Output

The output will be a video displaying the detected faces along with their estimated age and gender.

Example output:

Face Classification Output

Mood Classification

Files Included

  • mood_classifier.py: Python script for mood classification.
  • angry.mp4: Sample video for mood classification (angry).
  • surprised.jpg: Sample image for mood classification (surprised).

Getting Started

  1. Clone the Repository 
    git clone
cd AI-Classification/Mood-Classification
  2. Install Dependencies Install the required Python dependencies. 
    pip install -r requirements.txt
  3. Run Mood Classification 
    python mood_classifier.py

You will then be asked to choose your input type (image, video, or webcam). Enter image to classify the mood in the sample image provided (surprised.jpg), or enter video to classify the mood in a video file. You can also use your webcam for live testing.

The script will then display the detected mood in the image, video, or webcam stream and in the console.

All our classifiers will only stop when you press Q, ESC, or otherwise close the window.

Output

The output will be displayed the detected mood in the image, video, or webcam stream and in the console.

Example output:

Mood Classification Output

Character Classification

Files Included

  • ocr.py: Python script for character classification.
  • OIP.jpg: Sample JPEG image for character classification.
  • chars.jpg: Sample JPEG image for character classification.
  • chars.mp4: Sample video for character classification.
  • letters.mp4: Sample video for character classification.

Getting Started

  1. Clone the Repository

    git clone
cd AI-Classification/Character-Recognition

  2. Install the required Python dependencies.

    pip install -r requirements.txt

  3. Install Tessaract OCR

    • For Windows: Download and install the Tesseract OCR executable from the official Tesseract OCR website.
    • For Linux: Install Tesseract OCR using the package manager. 
      sudo apt-get install tesseract-ocr
    • For macOS: Install Tesseract OCR using Homebrew. 
      brew install tesseract
    • This is required for the OCR functionality to work. Also, when you install, note down the installation path of the Tesseract OCR executable. Replace the path in the pytesseract.pytesseract.tesseract_cmd variable in the ocr.py script with yours.
    • For example, if you installed Tesseract OCR in the default location on Windows, the path would be: 
      pytesseract.pytesseract.tesseract_cmd = r'C:\Program Files\Tesseract-OCR\tesseract.exe'

  4. Run Character Classification

    python ocr.py

    You will then be asked to choose your input type (image, video, or webcam). Enter image to classify the characters in the sample image provided (OIP.jpg), or enter video to classify characters in a video file. You can also use your webcam for live testing.

    image

The script will then display the detected characters in the image, video, or webcam stream.

All our classifiers will only stop when you press Q, ESC, or otherwise close the window.

Output

The output will display the class labels of the characters detected in the image along with the confidence scores.

Example output:

Character Classification Output

Flower Classification

Files Included

  • flower_classification.py: Python script for flower classification.
  • daisy.jpg: Sample JPEG image for flower classification (Daisy).
  • marigold.jpg: Sample JPEG image for flower classification (Marigold).
  • rose.mp4: Sample video for flower classification (Rose).

Getting Started

  1. Clone the Repository

    git clone https://github.com/hoangsonww/AI-Classification.git
cd AI-Classification/Flowers-Classification

  2. Install Dependencies Install the required Python dependencies.

    pip install -r requirements.txt

  3. Run Object Classification

    python flower_classification.py

You will then be asked to choose your input type (image, video, or webcam). Enter image to classify the flowers in the sample image provided (daisy.jpg), or enter video to classify flowers in a video file. You can also use your webcam for live testing.

All our classifiers will only stop when you press Q, ESC, or otherwise close the window.

Output

The output will display the class label of the flower detected in the image along with the confidence score.

Example output: Here are the sample image of Daisy flowers.

Flower Classification Output

Object Classification

Files Included

  • object_classification.py: Python script for object classification.
  • objects.jpg: Sample JPEG image for object classification.
  • objects.png: Sample PNG image for object classification.
  • balls.mp4: Sample video for object classification.
  • OIP.jpg: Sample image for object classification.

Getting Started

  1. Clone the Repository

    git clone https://github.com/hoangsonww/AI-Classification.git
cd AI-Classification/object_classification

  2. Install Dependencies Install the required Python dependencies.

    pip install -r requirements.txt

  3. Run Object Classification

    python object_classification.py

You will then be asked to choose your input type (image, video, or webcam). Enter image to classify the objects in the sample image provided (objects.jpg), or enter video to classify objects in a video file. You can also use your webcam for live testing.

Feel free to change the paths and other parameters in the script to suit your needs.

Note: All our classifiers will only stop when you press Q, ESC, or otherwise close the window.

Output

The output will display the class labels of the objects detected in the image along with the confidence scores. Or, if you choose to use your webcam, the output will display the class labels of the objects detected in the video stream. If you choose to use a video file, the output will be a video displaying the detected objects along with their class labels.

Example output:

Object Classification Output

Animal Classification

Files Included

  • animal_classification.py: Python script for animal classification.
  • cow.jpg: Sample JPEG image for animal classification (Cow).
  • ox.jpg: Sample JPEG image for animal classification (Ox).

Getting Started

  1. Clone the Repository

    git clone https://github.com/hoangsonww/AI-Classification.git
cd AI-Classification/Animals-Classification

  2. Install Dependencies Install the required Python dependencies.

    pip install -r requirements.txt

  3. Run Object Classification

    python animal_classification.py

The script will then ask you to choose your input type (image, video, or webcam). Enter image to classify the animals in the sample image provided (cow.jpg), or enter video to classify animals in a video file. You can also use your webcam for live testing.

All our classifiers will only stop when you press Q, ESC, or otherwise close the window.

Output

The output will display the class labels of the animals detected in the image along with the confidence scores.

Example output:

Animal Classification Output

Speech Recognition

Files Included

  • speech_classifier.py: Python script for speech recognition.
  • speech.mp4: Sample video file for speech recognition in a video context.
  • temp_audio.wav: Temp audio file (used by our AI) for speech recognition.

Getting Started

  1. Clone the Repository

    git clone https://github.com/hoangsonww/AI-Classification.git
cd AI-Classification/Speech-Recognition

  2. Install Dependencies Install the required Python dependencies.

    pip install -r requirements.txt

  3. Run Speech Recognition

     python speech_classifier.py

You will then be asked to choose your preferred input method (microphone or video). Enter microphone to use your microphone for live speech recognition, or enter video to use a video file for speech recognition.

Output

You will see the output of the speech recognition process in the console. The script will display the recognized speech from the audio input. The audio is processed in chunks and recognized in real-time. All our classifiers will stop when you press Q, ESC, or otherwise close the window.

Example output:

Speech Recognition Output

Special: Self-Trained Sentiment Classifier

In addition to the other pre-trained classifiers, this repository includes a special sentiment classifier that you can train yourself. The sentiment classifier is trained on a large dataset of tweets and can classify the sentiment of a sentence as positive, negative, or neutral. This is excellent for educational purposes and for understanding how sentiment analysis works.

Files Included

  • sentiment_classifier.py: Python script for sentiment classification.
  • train_model.py: Python script for training the sentiment classifier, which includes data preprocessing, model training, and evaluation.
  • sentiment_model.pkl: Trained sentiment classifier model.
  • vectorizer.pkl: Trained vectorizer for the sentiment classifier.
  • training.1600000.processed.noemoticon.csv: Training data for the sentiment classifier (Large file).
  • testdata.manual.2009.06.14.csv: Test data for the sentiment classifier.
  • test.csv: Sample test data for the sentiment classifier.
  • train.csv: Sample training data for the sentiment classifier.
  • generate_small_dataset.py: Python script for generating a small dataset from the large training data.
  • small_dataset.csv: Small dataset generated from the large training data.

Getting Started

  1. Clone the Repository

    git clone
cd AI-Classification/Sentiment-Analysis

  2. Install Dependencies: Install the required Python dependencies.

    pip install scikit-learn pandas numpy nltk tqdm joblib

  3. Pull the Large Training Data: The sentiment classifier is trained on a large dataset of tweets. The large training data is stored in a CSV file named training.1600000.processed.noemoticon.csv. This file is stored using Git LFS due to its large size. To pull the large training data, use the following command:

    git lfs install
git lfs pull

    • Alternatively, you can download the large training data from the Sentiment140 dataset website and place it in the Sentiment-Classifier directory. However, using Git LFS is recommended.

    • If you do not have Git LFS installed, remember to install it first. You can find instructions on how to install Git LFS on the official Git LFS website.

  4. Train the Sentiment Classifier Run the train_model.py script to train the sentiment classifier.

    python train_model.py

    • When running the script, you will be asked to choose the dataset size (small or large). Enter small to use the small dataset or large to use the large dataset. The script will then preprocess the training data, train the sentiment classifier, and save the trained model and vectorizer to disk.

    • However, if you choose small, the script will use the small dataset provided in the repository. In order to use it, be sure to run the generate_small_dataset.py script first to generate the small dataset from the large training data.

    python generate_small_dataset.py

    • Note: Training the sentiment classifier on the large dataset may take a long time and require significant computational resources. However, it is recommended since it provides better model accuracy.

    • Once again, if you are patient and have a good machine, you are encouraged use the large dataset to get a higher accuracy. Otherwise, use the small dataset for faster training.

    • This script will then preprocess the training data, train the sentiment classifier, and save the trained model and vectorizer to disk.

    • Additionally, it will output the expected accuracy, F1 score, and expected confidence level of the sentiment classifier. The higher these statistics are, the better the sentiment classifier will perform. Of course, this is highly dependent on the training dataset size and quality. Feel free to experiment with the training data and parameters to improve the sentiment classifier's performance.

  5. Run Sentiment Classification

    python sentiment_classifier.py

    You will then be asked to enter a sentence for sentiment classification. Enter a sentence, and the script will classify the sentiment of the sentence as positive, negative, or neutral, with a level of confidence.

Output

The output will display the sentiment classification of the input sentence. The sentiment classifier will classify the sentiment as positive, negative, or neutral.

Training Output Example:

Sentiment Classifier Training Output

Classification Output Example:

Sentiment Classifier Classification Output

Feel free to experiment with the sentiment classifier and test it with your own sentences and explore how powerful sentiment analysis can be!

Contact Information

For any questions or issues, please refer to the contact information below:

Future Work

  • Add more classifiers for various tasks such as emotion recognition, sentiment analysis, and more.
  • Refine existing classifiers and improve their accuracy and performance.
  • Add more sample images and videos for testing the classifiers.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Live Info Website

Feel free to visit the live demo and information website here.

It is a simple website that provides information about the classifiers in this repository.

This repository is a work in progress and under active development. If you have any suggestions or improvements, feel free to contribute to this repository. Thank you for visiting! 🚀

Created with ❤️ by Son Nguyen in 2024.