Transformer-Based Nonautoregressive Image Captioning via Guided Keyword Generation and Learnable Positional Encoding for IoT Devices | IEEE Journals & Magazine | IEEE Xplore