Emotion Detection with Deep Learning Models

Real-time and video-based emotion recognition using deep learning (VGG/LeNet/GoogLeNet). Supports live webcam streams and pre-recorded videos.

🔧 Basic Setup

1️⃣ Fork the repository

Fork this repository and enter the root folder (which should be named Emotion_Detection)

2️⃣ Create Conda Environment

conda create -n <venv_name> python=3.9.22
conda activate <venv_name>

3️⃣ Install Dependencies

Core Requirements:

• OpenCV 4.5+
• TensorFlow 2.6+
• Other packages from requirements.txt

pip install -r requirements.txt

💡 Conda Tip: For GPU support, install TensorFlow with Conda:

conda install -c conda-forge tensorflow-gpu

 

🚀 Quick Start Guide

Prerequisites:

• You are in the project root directory
• Environment is set up (requirements.txt installed)

---

 

1️⃣ Built Dataset

python emo_rec/build_dataset.py

📁 Output:

• The training, validation, and testing datasets will be stored under datasets/fer2013/hdf5

---

 

2️⃣ Train a Model

Option A: Specify the Model You Want to Train

python emo_rec/train_emotion_detector.py -m <model_name>

Available Models:

• emotionvggnet (default)
• lenet
• minigooglenet
• minivggnet
• shallownet

Option B: Train Default Model (emotionvggnet)

python emo_rec/train_emotion_detector.py

📁 Outputs:

• Trained models: emo_rec/built_models/
• Training logs: emo_rec/training_logs/

---

 

3️⃣ Test a Model

python emo_rec/test_emotion_detector.py -m <model_name>

❗ Requirements:

• Model must be trained first (Step 1)
• -m flag is mandatory

Example:

python emo_rec/test_emotion_detector.py -m minivggnet

---

4️⃣ Run Emotion Detection

Option A: Process Video File

python emo_rec/run_emotion_detector.py -m <model_name> -v <video_path>

Example (to run the demo video already provided):

python emo_rec/run_emotion_detector.py -m emotionvggnet -v emo_rec/video/example.mp4

Option B: Real-Time Webcam

python emo_rec/run_emotion_detector.py -m <model_name>

🖥️ Controls:

1. Select the camera window
2. Press:
   • Mac: Cmd + Q
   • Windows: Ctrl + Q

Example:

python emo_rec/run_emotion_detector.py -m lenet

🔴 Important:

• Model must be trained first (Step 1)
• -m <model_name> is always required
• Uses default camera device