|
|
|
_base_ = 'coco_instance.py' |
|
dataset_type = 'LVISV05Dataset' |
|
data_root = 'data/lvis_v0.5/' |
|
data = dict( |
|
samples_per_gpu=2, |
|
workers_per_gpu=2, |
|
train=dict( |
|
_delete_=True, |
|
type='ClassBalancedDataset', |
|
oversample_thr=1e-3, |
|
dataset=dict( |
|
type=dataset_type, |
|
ann_file=data_root + 'annotations/lvis_v0.5_train.json', |
|
img_prefix=data_root + 'train2017/')), |
|
val=dict( |
|
type=dataset_type, |
|
ann_file=data_root + 'annotations/lvis_v0.5_val.json', |
|
img_prefix=data_root + 'val2017/'), |
|
test=dict( |
|
type=dataset_type, |
|
ann_file=data_root + 'annotations/lvis_v0.5_val.json', |
|
img_prefix=data_root + 'val2017/')) |
|
evaluation = dict(metric=['bbox', 'segm']) |
|
|
