Link of the Paper: https://arxiv.org/pdf/1504.06692.pdf
Innovations:
- The authors propose the Novel Visual Concept learning from Sentences ( NVCS ) task. In this task, methods need to learn novel concepts from sentence descriptions of a few images.
- The authors propose a transposed weight sharing scheme, which not only improves performance on image captioning, but also makes the model more suitable for the novel concept learning task.
- Three novel concept datasets are constructed for NVCS task.
Improvements:
- The method proposed is based on m-RNN and the authors provide several inprovements.
General Points: