Welding　is　integral　to　modern　manufacturing,　yet　the　complex　process　often　leads　to　defects,　impacting　the　quality　of　the　final　product.　Recent　advances　in　deep　learning,　particularly　Convolutional　Neural　Networks　(CNNs),　have　shown　remarkable　results　in　applications　like　defect　recognition.　This　study　evaluated　AlexNet,　ResNet-18,　ResNet50,　ResNet-101,　MobileNet-v2,　ShuffleNet,　and　SqueezeNet　for　their　effectiveness　in　identifying　welding　defects,　using　accuracy,　precision,　sensitivity,　specificity,　and　F-score　as　metrics.　The　dataset　covered　defects　like　cracks,　lack　of　penetration,　porosity,　and　a　no-defect　class.　Our　analysis　shows　that　most　of　these　architectures　deliver　promising　results　in　accuracy,　sensitivity,　specificity,　precision,　and　F1-score,　highlighting　their　potential　in　defect　recognition.　©　The　2024　International　Conference　on　Artificial　Life　and　Robotics　(ICAROB2024).