Is it possible to combine text prompts? I.e. to detect multiple types of objects in one forward pass.

[dempsey-ryan]

e.g., detecting both trees and buildings in a satellite image. I tried inputting both LangSam.predict(text_prompt="tree, building") and LangSam.predict(text_prompt="building, tree") and they yield fairly differing results. I'm wondering if there is a method to detect multiple object prompts without an inherent hierarchy in the prompts due to the order of the text.

[giswqs]

Multiple types are not supported yet. I don't have a timeline when it will be implemented. Contributions are welcome.