The scarcity of comprehensive datasets in surveillance, identification, image retrieval systems, and healthcare poses a significant challenge for researchers in exploring new methodologies and advancing knowledge in these respective fields. Furthermore, the need for full-body image datasets with detailed attributes like height, weight, age, and gender is particularly significant in areas such as fashion industry analytics, ergonomic design assessment, virtual reality avatar creation, and sports performance analysis. To address this gap, we have created the 'Celeb-FBI' dataset which contains 7,211 full-body images of individuals accompanied by detailed information on their height, age, weight, and gender. Following the dataset creation, we proceed with the preprocessing stages, including image cleaning, scaling, and the application of Synthetic Minority Oversampling Technique (SMOTE). Subsequently, utilizing this prepared dataset, we employed three deep learning approaches: Convolutional Neural Network (CNN), 50-layer ResNet, and 16-layer VGG, which are used for estimating height, weight, age, and gender from human full-body images. From the results obtained, ResNet-50 performed best for the system with an accuracy rate of 79.18% for age, 95.43% for gender, 85.60% for height and 81.91% for weight.