The main consideration when designing a task to get more responses is whether or not the contributor will be able to complete job 2 immediately after job 1. From a contributor's point of view, image annotation is a relatively "high-effort" task, so if you ask multiple image annotation questions per job, the contributor is more likely to quit halfway, give unclear responses, or stop after completing one job.
To get more responses, try streamlining the process for contributors by only asking one image annotation question per job. If you have more than one question about an image, all you need to do is break them up into separate jobs. If you'd like to include more than one question in the same job, try putting the image annotation question at the beginning so that when contributors view the image, their focus is sharp enough to give the most accurate response possible.