Hello, thank you for the interest of our work.
Indeed, we use idea of prompt tuning in our work, but we leverage it as a kind of conditioning method for our Transformer-based image codec to achieve variable-rate and ROI coding. The whole system (including prompts and codec itself) is trained end-to-end. More details can be found in our paper.
Hello, thank you for the interest of our work. Indeed, we use idea of prompt tuning in our work, but we leverage it as a kind of conditioning method for our Transformer-based image codec to achieve variable-rate and ROI coding. The whole system (including prompts and codec itself) is trained end-to-end. More details can be found in our paper.