Emo0.1 – Facial Emotion Recognition (FER2013)

Emo0.1 is a lightweight TensorFlow/Keras model fine-tuned on VGG16 to classify 7 facial emotions from the FER2013 dataset.

Model Details

• Input: 48x48 RGB face images
• Output: ['angry', 'disgusted', 'fearful', 'happy', 'neutral', 'sad', 'surprised']
• Framework: TensorFlow 2.x
• Training: 30 epochs with Adam (lr=0.0001)

Quick Start

Installation

pip install tensorflow opencv-python

Load Model

from huggingface_hub import hf_hub_download
from tensorflow.keras.models import load_model

model_path = hf_hub_download(repo_id="shivampr1001/Emo0.1", filename="Emo0.1.h5")
model = load_model(model_path)

Predict from Image

import cv2
import numpy as np

def predict_emotion(img_path):
    img = cv2.imread(img_path)
    img = cv2.resize(img, (48,48)).astype('float32')/255
    pred = model.predict(np.expand_dims(img, axis=0))[0]
    return ['angry','disgusted','fearful','happy','neutral','sad','surprised'][np.argmax(pred)]

File	Purpose
facial_EmotionClassifer.h5	Pre-trained Keras model
RealTimeClassification.py	Webcam-based emotion prediction
predictBY_img.py	Image-based emotion prediction
haarcascade_frontalface_default.xml	Haar cascade for face detection

License Apache 2.0 © Shivam Prasad