We have produced an automatic welding defect detection system, which is a computer software application, using Radiation Sourced images. Our software can take the image of a pipe with its welding defects and compare it to an optimal database where the “perfect” constitution of the image is stored, and indicate deficiencies in real-time, using Machine Learning Algorithms (ML). Our proposed method for automatic detection and classification of faults and defects in the radiographic images of welded joints is attained through an exposure technique of double-wall image (DWDI).