problem with registering custom dataset about centernet2 HOT 7 CLOSED

i am no longer using colab i am doing it on local now but I can help you with it you just have to go to train_net.py and add these line from detectron2.data.datasets import register_coco_instances
register_coco_instances("my_dataset_train", {}, "/content/results/yolo2coco/train.json", "/content/results/yolo2coco/train")
register_coco_instances("my_dataset_val", {}, "/content/results/yolo2coco/test.json", "/content/results/yolo2coco/test")
register_coco_instances("my_dataset_test", {}, "/content/results/yolo2coco/test.json", "/content/results/yolo2coco/test")
make changes according to your data and add my_dataset_train and my_dataset_test in config file datasets key as mentioned above

from centernet2.

@gauravlochab Thank you for the quick response

from centernet2.

let me know if you get good results :p

from centernet2.

Hi,
From the error log, it seems the config file is not in the correct format. Did you use TRAIN = ('my_dataset_train',) instead of TRAIN: ('my_dataset_train',)?

from centernet2.

thanks for the quick reply yes that was the issue, I am training on my custom data on colab with one gpu. could you tell me the best way to fine-tune this on my data what all hyperparameters i should be looking for while training
21:08:40 iter: 186080 total_loss: 2.043 loss_cls_stage0: 0.2828 loss_box_reg_stage0: 0.02241 loss_cls_stage1: 0.2958 loss_box_reg_stage1: 0.02742 loss_cls_stage2: 0.2986 loss_box_reg_stage2: 0.02829 loss_centernet_loc: 0.6763 loss_centernet_agn_pos: 0.3532 loss_centernet_agn_neg: 0.02419 time: 0.8431 data_time: 0.0036 lr: 0.039999
last iteration looks like this. Thanks again

from centernet2.

@gauravlochab Could you share the notebook you are working on? I'm stuck at the data register.

from centernet2.

https://blog.csdn.net/Carlsummer/article/details/116711708

from centernet2.