During my internship at Centre for Airborne Systems (CABS), DRDO, I had the opportunity to work on a gender classification model aimed at integration into NVIDIA DeepStream for video analytics. This model has been trained to classify genders with a remarkable accuracy of 96%.
The project aimed to develop a robust gender classification model suitable for real-time applications, particularly for deployment within video analytics pipelines. Key aspects of the project include:
Data Collection and Preprocessing: Curating a diverse dataset of facial images representing different genders, followed by preprocessing steps to enhance model performance.
Model Development: Designing and training a convolutional neural network (CNN) model capable of accurately classifying gender from facial images.
Integration with Video Analytics Pipelines: Adapting the trained model to seamlessly integrate with existing video analytics frameworks, enabling real-time gender classification within video streams.
Performance Evaluation: Rigorous testing and evaluation of the integrated model to ensure high accuracy and reliability in gender classification tasks.
-
Successfully trained a gender classification model achieving an impressive accuracy of 96%.
-
Integrated the model into video analytics pipelines, facilitating real-time gender classification within video streams.
-
Contributed to the advancement of AI capabilities within defense and security applications, enabling enhanced surveillance and analysis capabilities.
Model Training Code: Contains scripts and notebooks used for data preprocessing, model training, and evaluation.
Documentation: Detailed documentation covering model architecture, training process, integration steps, and usage guidelines.