This study explored using computer vision and machine learning to predict body measurements from 2D images for custom garment production. Mask R-CNN and Keypoint R-CNN algorithms extracted features from images, which were then used to train ML models, including Decision Tree, Random Forest, Gradient Boosting, and BPANN, among others. The methodology involved using Mask R-CNN and Keypoint R-CNN for feature extraction, followed by training various models to establish the relationship between features and body measurements. The models were evaluated to identify the best one for generating custom garment patterns. Results indicated that Random Forest is the bestperforming model, with its best result in the shoulder width body part with 0.79 R2 score, 0.25 MSE, 0.44 MAE, 0.50 RMSE, and a Max Error of 0.90, demonstrating superior accuracy and consistency, while Gradient Boosting also performed well but with higher variability. BPANN was the lowest-performing model, highlighting the need for careful model selection. The study aimed to streamline obtaining accurate body dimensions, reducing the time and invasiveness of traditional methods. The findings underscore the importance of thorough model evaluation to develop a reliable system for predicting body measurements, enhancing custom garment production. To build upon these findings and address limitations, future research should refine models, explore advanced techniques, and gather a larger, more diverse dataset to improve accuracy and robustness. Expanding garment pattern options will enhance personalization. Additionally, refining feature extraction techniques will improve model performance, leading to more accurate measurements and better-fitting custom garments, making the system more reliable and effective.