VQA input is not the dependency task's result.

[CLNET-pixel]

from the log, u can see vqa's input is not objection detection's output("inference result": {"generated image with predicted box": "/images/7e4e.jpg"), but its input... I checked the code, there is no key named "generated image with predicted box" in dependency task results get.

[CLNET-pixel]

Is that a code bug? when detected the results of dependency tasks, no considering these two new keys.

[tricktreat]

@CLNET-pixel Thanks for the suggestion. We will fix this bug ASAP.