COVID-19　has　devastated　the　global　healthcare　system　as　well　as　the　world　economy　with　more　than　600　million　confirmed　cases　and　6　million　deaths　globally.　A　timely　and　accurate　diagnosis　of　the　disease　plays　a　vital　role　in　the　treatment　and　preventative　spread　of　disease.　Recently,　deep　learning　such　as　Convolutional　Neural　Networks　(CNNs)　have　achieved　extraordinary　results　in　many　applications　such　as　medical　classifications.　This　work　focuses　on　investigating　the　performance　of　nine　state-of-the-art　architectures:　Alexnet,　Googlenet,　Inception-v3,　Mobilenet-v2,　Resnet-18,　Resnet-50,　Shufflenet,　Squeezenet　and　Resnet-50　RCNN　for　COVID-19　classification　by　comparing　with　performance　metrics　such　as　accuracy,　precision,　sensitivity,　specificity　and　F-score.　The　datasets　considered　in　current　study　are　divided　into　three　different　classes　namely　Normal　Chest　X-Rays　(CXRs),　Pneumonia　patient　CXR　and　COVID-19　patient　CXR.　The　results　achieved　shows　that　Resnet-50　RCNN　achieved　an　accuracy,　precision,　sensitivity,　specificity　and　F-score　of　95.67%,　95.71%,　95.67%,　97.84%　and　95.67%　respectively.　©　The　2023　International　Conference　on　Artificial　Life　and　Robotics　(ICAROB2023).