How to Fine-tune with pretrained weights from Model Zoo on custom datasets?

[guoqsGary]

How to Fine-tune with pretrained weights from Model Zoo on custom datasets?
change the init_checkpoint where: train.init_checkpoint = "detectron2://ImageNetPretrained/torchvision/R-50.pkl" ?

[rentainhe]

How to Fine-tune with pretrained weights from Model Zoo on custom datasets? change the init_checkpoint where: train.init_checkpoint = "detectron2://ImageNetPretrained/torchvision/R-50.pkl" ?

Yes, you can simply update config by setting train.init_checkpoint to the pretrained weights then train on your own dataset

[guoqsGary]

Thanks! I want to ask an other question.
How to visualize predictions on my own datasets?
I have regist my dataset in "coco_detr.py" but still got the error "KeyError: "Dataset 'my_dataset_test' is not registered!"

[rentainhe]

Thanks! I want to ask an other question. How to visualize predictions on my own datasets? I have regist my dataset in "coco_detr.py" but still got the error "KeyError: "Dataset 'my_dataset_test' is not registered!"

Would you like to share the whole content of you config, which may be helpful for us to check the issue

[guoqsGary]

from omegaconf import OmegaConf

import detectron2.data.transforms as T
from detectron2.config import LazyCall as L
from detectron2.data import (
build_detection_test_loader,
build_detection_train_loader,
get_detection_dataset_dicts,
)
from detectron2.data.datasets import register_coco_instances
from detectron2.evaluation import COCOEvaluator

from detrex.data import DetrDatasetMapper

dataloader = OmegaConf.create()

register_coco_instances("my_dataset_train", {}, '/home/workspace/datasets/annotations/instances_train2017.json', '/home/workspace/datasets/train2017')
register_coco_instances("my_dataset_test", {}, '/home/workspace/datasets/annotations/instances_val2017.json', '/home/workspace/datasets/val2017')

dataloader.train = L(build_detection_train_loader)(
dataset=L(get_detection_dataset_dicts)(names="my_dataset_train"),
mapper=L(DetrDatasetMapper)(
augmentation=[
L(T.RandomFlip)(),
L(T.ResizeShortestEdge)(
short_edge_length=(480, 512, 544, 576, 608, 640, 672, 704, 736, 768, 800),
max_size=1333,
sample_style="choice",
),
],
augmentation_with_crop=[
L(T.RandomFlip)(),
L(T.ResizeShortestEdge)(
short_edge_length=(400, 500, 600),
sample_style="choice",
),
L(T.RandomCrop)(
crop_type="absolute_range",
crop_size=(384, 600),
),
L(T.ResizeShortestEdge)(
short_edge_length=(480, 512, 544, 576, 608, 640, 672, 704, 736, 768, 800),
max_size=1333,
sample_style="choice",
),
],
is_train=True,
mask_on=False,
img_format="RGB",
),
total_batch_size=4,
num_workers=4,
)

dataloader.test = L(build_detection_test_loader)(
dataset=L(get_detection_dataset_dicts)(names="my_dataset_test", filter_empty=False),
mapper=L(DetrDatasetMapper)(
augmentation=[
L(T.ResizeShortestEdge)(
short_edge_length=800,
max_size=1333,
),
],
augmentation_with_crop=None,
is_train=False,
mask_on=False,
img_format="RGB",
),
num_workers=4,
)

dataloader.evaluator = L(COCOEvaluator)(
dataset_name="${..test.dataset.names}",
)

Thanks! I want to ask an other question. How to visualize predictions on my own datasets? I have regist my dataset in "coco_detr.py" but still got the error "KeyError: "Dataset 'my_dataset_test' is not registered!"

Would you like to share the whole content of you config, which may be helpful for us to check the issue