I try to use the api of obbdetection for training (as used in mmdetection) but I got this error: TypeError: myRegister: init() got an unexpected keyword argument 'task'
Here my script for for configuration and training:
#################################################################
def configuration():
backbone = 'obb/oriented_rcnn/faster_rcnn_orpn_r50_fpn_1x_dota10.py'
weight = 'faster_rcnn_orpn_r50_fpn_1x_dota10_epoch12.pth'
cfg = Config.fromfile(f"{root}/OBBDetection/configs/{backbone}")
cfg.load_from = f"{root}/OBBDetection/ckpt/{weight}"
# Modify dataset type and path
cfg.dataset_type = 'myDataLoader'
cfg.data_root = ''
cfg.data.train.type = 'myRegister'
cfg.data.train.data_root = ''
cfg.data.train.ann_file = f'{annotations_dir}/train_dicts_roof_plan.json'
cfg.data.train.img_prefix = ''
cfg.data.test.type = 'myRegister'
cfg.data.test.data_root = ''
cfg.data.test.ann_file = f'{annotations_dir}/val_dicts_roof_plan.json'
cfg.data.test.img_prefix = ''
# modify num classes of the model in box head
cfg.model.roi_head.bbox_head.num_classes = len(CLASSES)
# Set up working dir to save files and logs.
cfg.work_dir = work_dir
# Toatal epochs
cfg.total_epochs = total_epochs
# Change the evaluation metric since we use customized dataset.
cfg.evaluation = 'mAP'
cfg.total_epochs = 50
cfg.optimizer.lr = 5e-4
cfg.data.samples_per_gpu = 4
cfg.data.workers_per_gpu = 4
return cfg
I try to use the api of obbdetection for training (as used in mmdetection) but I got this error: TypeError: myRegister: init() got an unexpected keyword argument 'task' Here my script for for configuration and training: ################################################################# def configuration():
################################################################# def my_training(cfg):
###########################################################