Abstract: The Convolutional Neural Network (CNN) has shown remarkable versatility across various applications. One such emerging area of research is age range estimation using CNN, which finds applications in numerous domains and continues to be a state-of-the-art field for investigation, aiming to enhance estimation accuracy.

In our proposed work, we leverage a deep CNN model to identify people's age ranges. Initially, we employ the MTCNN technique to extract only face images from the dataset, discarding unnecessary features unrelated to the face. To improve the model's performance, we utilize data augmentation through random crop techniques.

Our approach also incorporates transfer learning, where a pretrained face recognition model, VGG-Face, serves as the foundation for building our age range identification model. The effectiveness of our work is evaluated on the Adience Benchmark, where the performance on the test set significantly surpasses existing state-of-the-art methods.

Overall, our research demonstrates the prowess of CNN-based age range estimation, showcasing promising results and contributing to the advancement of this domain.

Keywords: Computer Vision, Image Detection, Feature Matching, MTCNN, NumPy, OpenCV, PyTorch , FaceNet , P-Net, O-Net , R-Net , NMS , Face-Alignment.

| DOI: 10.17148/IARJSET.2023.107125