mmyolo/configs/rtmdet/rotated/rtmdet-r_tiny_fast_1xb8-36e_dota.py
Yanyi Liu 4b996f10a5
Rotated object detection and RTMDet-R (#513)
* init

* add cfg

* update align

* update

* fix regularize box

* fix comment

* update config

* remove ckpt

* update

* make mmrotate optional

* fix doc

* add mmrotate req

* support large_demo with rbbox

* add ut

* update

* add doc v01

* update doc

* fix doc

* update

* update

* update readme

* update comments

* fix

* fix doc

* fix doc

* fix

* update

* update

* fix large

* update doc

* update readme

* fix config

* fix configs

* inprove

* update doc

* update assigner

* update ut

* remove rdsl assigner

* rename aug config

* speedup ut

* add comment

* fix data root

* remove doc

* remove empty folder

* add docs

* rename configs

* fix readme

* fix readme

* fix configs

* revert

* fix name

* fix table

* fix doc link

* fix doc link

* update

* update

* update

* Refactor dota splits

* add shapely

* fix typo

* fix ci

* change

* fix type

* uppdata link

* uppdata link

* add some comment

* update

---------

Co-authored-by: huanghaian <huanghaian@sensetime.com>
2023-03-02 10:27:46 +08:00

39 lines
1.4 KiB
Python

_base_ = './rtmdet-r_l_syncbn_fast_2xb4-36e_dota.py'
checkpoint = 'https://download.openmmlab.com/mmdetection/v3.0/rtmdet/cspnext_rsb_pretrain/cspnext-tiny_imagenet_600e.pth' # noqa
# ========================modified parameters======================
deepen_factor = 0.167
widen_factor = 0.375
# Batch size of a single GPU during training
train_batch_size_per_gpu = 8
# Submission dir for result submit
submission_dir = './work_dirs/{{fileBasenameNoExtension}}/submission'
# =======================Unmodified in most cases==================
model = dict(
backbone=dict(
deepen_factor=deepen_factor,
widen_factor=widen_factor,
init_cfg=dict(checkpoint=checkpoint)),
neck=dict(deepen_factor=deepen_factor, widen_factor=widen_factor),
bbox_head=dict(head_module=dict(widen_factor=widen_factor)))
train_dataloader = dict(batch_size=train_batch_size_per_gpu)
# Inference on test dataset and format the output results
# for submission. Note: the test set has no annotation.
# test_dataloader = dict(
# dataset=dict(
# data_root=_base_.data_root,
# ann_file='', # test set has no annotation
# data_prefix=dict(img_path=_base_.test_data_prefix),
# pipeline=_base_.test_pipeline))
# test_evaluator = dict(
# type='mmrotate.DOTAMetric',
# format_only=True,
# merge_patches=True,
# outfile_prefix=submission_dir)