Description
Hi, I am trying to train Panoptic FPN model on custom dataset (COCO Panoptic format). For my dataset num_stuff_classes=4 and num_things_classes=13. Running on Linux platform with CPU under conda environment. My config file is as follows:
###########
the new config inherits the base configs to highlight the necessary modification
base = './panoptic_fpn/panoptic-fpn_r50_fpn_1x_coco.py'
METAINFO = {
'classes':
('a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n', 'o', 'p', 'q'),
'thing_classes':
('b', 'c', 'd', 'f', 'g', 'h', 'i', 'j', 'k', 'n', 'o', 'p', 'q'),
'stuff_classes':
('a', 'e', 'l', 'm'),
'palette':
[(220, 20, 60), (119, 11, 32), (0, 0, 142), (0, 0, 230), (106, 0, 228),
(0, 60, 100), (0, 80, 100), (0, 0, 70), (0, 0, 192), (250, 170, 30),
(100, 170, 30), (220, 220, 0), (175, 116, 175), (250, 0, 30),
(165, 42, 42), (255, 77, 255), (0, 226, 252), (50,150,10)]
}
explicitly over-write all the num_classes field from default
model = dict(
roi_head=dict(
bbox_head=dict(
num_classes=17),
mask_head=dict(
num_classes=17),),
semantic_head=dict(
num_stuff_classes=4,
num_things_classes=13,),
panoptic_fusion_head=dict(
num_stuff_classes=4,
num_things_classes=13))
update panoptic annotation files location and METAINFO
data_root = 'DACC/'
dataset_type = 'CocoPanopticDataset'
train_cfg = dict(max_epochs=4, type='EpochBasedTrainLoop', val_interval=4)
train_dataloader = dict(
dataset=dict(
type=dataset_type,
ann_file='annotations/Panoptic_train.json',
metainfo=METAINFO,
data_root=data_root,
data_prefix=dict(
img='train/', seg='annotations/panoptic_train/')
),
num_workers=1
)
val_dataloader = dict(
dataset=dict(
type=dataset_type,
ann_file='annotations/Panoptic_val.json',
metainfo=METAINFO,
data_root=data_root,
data_prefix=dict(
img='val/', seg='annotations/panoptic_val/')
),
num_workers=1
)
test_dataloader = dict(
dataset=dict(
type=dataset_type,
ann_file='annotations/Panoptic_val.json',
metainfo=METAINFO,
data_root=data_root,
data_prefix=dict(
img='val/', seg='annotations/panoptic_val/')
),
num_workers=1
)
val_evaluator = dict(
ann_file='DACC/annotations/Panoptic_val.json',
backend_args=None,
seg_prefix='DACC/annotations/panoptic_val/',
type='CocoPanopticMetric')
test_evaluator = dict(
ann_file='DACC/annotations/Panoptic_val.json',
backend_args=None,
seg_prefix='DACC/annotations/panoptic_val/',
type='CocoPanopticMetric')
load_from = '/home/kou/anaconda3/envs/openmmlab/mmdetection/checkpoints/panoptic_fpn_r50_fpn_1x_coco_20210821_101153-9668fd13.pth'
##############
Now, my training is getting killed without any error message, screenshot attached. I have tried to adjust num_classes under roi_head(bbox_head, mask_head) to 13 and 17 both separately , but it doesn't work.
For num_classes=13 it throws "IndexError: index 13 is out of bounds for dimension 1 with size 13" , screenshot attached below.
Can someone help me point out my mistake in configuration settings?
Thanking you _/_
