How to get instance/semantic segmentation feature

[h-peng17]

Hi~
We can get object detection feature by running the model step by step like https://detectron2.readthedocs.io/en/latest/tutorials/models.html. But how can I get instance/semantic segmentation feature from models? The predicted instances has pred_masks, how can I use it? Wish for your reply~

[deepaksinghcv]

In https://detectron2.readthedocs.io/en/latest/tutorials/models.html#model-output-format, it is mentioned that in inference mode the model will output a list[dict] for each image with appropriate keys to access pred_masks
@ppwwyyxx Kindly correct me if I'm wrong.