@deepcs233
Hello, thank you for open-sourcing your work. I am working on obstacle avoidance for water surface unmanned boats and am interested in applying your framework to my research. Given that constraints on water surfaces are generally less than those on roads, I want to know if it's feasible to fine-tune LMDrive with a dataset of water surface images that I collect to achieve obstacle avoidance on water. If feasible, would it be necessary to fine-tune both the visual encoder and the command instructions?
@deepcs233 Hello, thank you for open-sourcing your work. I am working on obstacle avoidance for water surface unmanned boats and am interested in applying your framework to my research. Given that constraints on water surfaces are generally less than those on roads, I want to know if it's feasible to fine-tune LMDrive with a dataset of water surface images that I collect to achieve obstacle avoidance on water. If feasible, would it be necessary to fine-tune both the visual encoder and the command instructions?