IMS_ PER_ BATCH: about geoshift HOT 5 CLOSED

Hello,

Thanks for your interest in our work. Yes, you can change it to 1 but in all of our experiments, we used batch_size as mentioned in the respective yaml files. Additionally, you may have to increase the max iterations when batch size is lowered to 1.

from geoshift.

Thank you for your reply. To be honest, I am a first-year graduate student from China.
In this regard, I am a complete novice.Just now, I completed the training for the third step. I'm wondering if there are sentences similar to "python demo.py --config-file D:/learning/detectron2-main/configs/COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml --input test5.jpg --opts MODEL.WEIGHTS detectron2://COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x/137849600/model_final_f10217.pkl" that can be generated using the trained model.
This question might be basic, but I still look forward to your response.

from geoshift.

You can use --eval-only flag to the commands python <train-filename> --eval-only --config-file <config-filename> MODEL.WEIGHTS <path to the weights> . This command will start the evaluation of the test dataset. You can look into this and modify it to evaluate only one image.

from geoshift.

Thank you for your reply, and I wish you smooth work and good health!

from geoshift.

Thank you and same to you!

from geoshift.