what image size has been used for the final segmentation training?

[seyeeet]

what image size has been used for the final segmentation training when we want to training the model on all the acquired labels?
Also, can you please clarify if we need to use the coco weights in the final training for the initialization or not? I cannot get the reproduce results for the training stage at for the final stage, so any clarification will be very useful
Thanks