Implemented an Encoder-Decoder model in TensorFlow, where ResNet-50 extracts features from the VizWiz-Captions image dataset and a GRU with Bahdanau attention generates captions.