This paper introduces AM-SegNet, a lightweight deep learning model designed for semantic segmentation and feature quantification of high-resolution X-ray images in additive manufacturing (AM) processes. The model is trained on a large-scale database of over 10,000 pixel-labeled X-ray images, enabling efficient and accurate segmentation. AM-SegNet incorporates a lightweight convolution block and a custom attention mechanism, achieving high accuracy (around 96%) and fast processing speed (<4 ms per frame). The model is capable of segmenting and quantifying critical features such as keyholes and pores in AM processes, and has been applied to various advanced manufacturing techniques, including laser powder bed fusion (LPBF) and high-pressure die casting (HPDC). The model's performance is validated across different AM experiments, demonstrating its effectiveness in real-time segmentation and quantification of X-ray images. The study also highlights the potential of AM-SegNet to improve the efficiency and accuracy of data processing in manufacturing and imaging domains, enabling deeper insights into the physical processes involved in AM. The proposed method offers a generalizable solution for AM X-ray image segmentation and feature quantification, with the potential to enhance the reliability and consistency of AM processes.This paper introduces AM-SegNet, a lightweight deep learning model designed for semantic segmentation and feature quantification of high-resolution X-ray images in additive manufacturing (AM) processes. The model is trained on a large-scale database of over 10,000 pixel-labeled X-ray images, enabling efficient and accurate segmentation. AM-SegNet incorporates a lightweight convolution block and a custom attention mechanism, achieving high accuracy (around 96%) and fast processing speed (<4 ms per frame). The model is capable of segmenting and quantifying critical features such as keyholes and pores in AM processes, and has been applied to various advanced manufacturing techniques, including laser powder bed fusion (LPBF) and high-pressure die casting (HPDC). The model's performance is validated across different AM experiments, demonstrating its effectiveness in real-time segmentation and quantification of X-ray images. The study also highlights the potential of AM-SegNet to improve the efficiency and accuracy of data processing in manufacturing and imaging domains, enabling deeper insights into the physical processes involved in AM. The proposed method offers a generalizable solution for AM X-ray image segmentation and feature quantification, with the potential to enhance the reliability and consistency of AM processes.