-
Notifications
You must be signed in to change notification settings - Fork 5
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
larger size of femtoDet in coco dataset
- Loading branch information
Showing
4 changed files
with
696 additions
and
0 deletions.
There are no files selected for viewing
174 changes: 174 additions & 0 deletions
174
configs/femtoDet_large_coco/femtodet_large_coco_0stage.py
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,174 @@ | ||
| #dataset | ||
| # classes = ('aeroplane', 'bicycle', 'bird', 'boat', 'bottle', 'bus', 'car', 'cat', 'chair', 'cow', 'diningtable', | ||
| # 'dog', 'horse', 'motorbike', 'person', 'pottedplant', 'sheep', 'sofa', 'train', 'tvmonitor') | ||
| data_root = './data/coco/' | ||
| #model | ||
| img_scale = (640, 640) | ||
| widen_factor=1.0 | ||
| num_last_epochs=0 | ||
| max_epochs=300 | ||
| test_img_scale=(416,416) | ||
| pretrain_weight='./weights/femto_large_imagenet1k.pth' | ||
| load_from = None | ||
| resume_from = None | ||
| #more information | ||
| optimizer = dict( | ||
| type='SGD', | ||
| lr=0.01, | ||
| momentum=0.90, | ||
| weight_decay=0.0005, | ||
| nesterov=True, | ||
| paramwise_cfg=dict(norm_decay_mult=0.0, bias_decay_mult=0.0)) | ||
| optimizer_config = dict(grad_clip=None) | ||
| lr_config = dict( | ||
| policy='YOLOX', | ||
| warmup='exp', | ||
| by_epoch=False, | ||
| warmup_by_epoch=True, | ||
| warmup_ratio=1, | ||
| warmup_iters=5, | ||
| num_last_epochs=num_last_epochs, | ||
| min_lr_ratio=0.05) | ||
| runner = dict(type='EpochBasedRunner', max_epochs=max_epochs) | ||
| checkpoint_config = dict(interval=10) | ||
| log_config = dict(interval=50, hooks=[dict(type='TextLoggerHook')]) | ||
| custom_hooks = [ | ||
| dict(type='FemtoDetModeSwitchHook', num_last_epochs=num_last_epochs, priority=48, all_unfreeze_after_this_epoch=50), | ||
| dict(type='SyncNormHook', num_last_epochs=num_last_epochs, interval=10, priority=48), | ||
| dict( | ||
| type='ExpMomentumEMAHook', | ||
| resume_from=None, | ||
| momentum=0.0001, | ||
| priority=49) | ||
| ] | ||
| find_unused_parameters=True | ||
| dist_params = dict(backend='nccl') | ||
| log_level = 'INFO' | ||
| workflow = [('train', 1)] | ||
| opencv_num_threads = 0 | ||
| mp_start_method = 'fork' | ||
| default_channels=[32, 96, 320] | ||
| img_norm_cfg = dict(mean=[123.675, 116.28, 103.53], std=[58.82, 58.82, 58.82], to_rgb=True) | ||
| neck_in_chanels = [int(ch*widen_factor) for ch in default_channels] | ||
| headfeat_channel = 64 | ||
| model = dict( | ||
| type='FemtoDet', | ||
| input_size=img_scale, | ||
| random_size_range=(10, 20), | ||
| random_size_interval=10, | ||
| backbone=dict( | ||
| type='FemtoNet', | ||
| widen_factor=widen_factor, | ||
| diff_conv=True, | ||
| out_indices=(2, 4, 6), | ||
| act_cfg=dict(type='ReLU'), | ||
| init_cfg=dict( | ||
| type='Pretrained', | ||
| prefix='backbone', | ||
| checkpoint=pretrain_weight)), | ||
| neck=dict( | ||
| type='SharedNeck', | ||
| in_channels=neck_in_chanels, | ||
| out_channels=headfeat_channel, | ||
| fixed_size_idx=1, | ||
| add=True, | ||
| norm_cfg=dict(type='BN'), | ||
| act_cfg=dict(type='ReLU'), | ||
| num_outs=1), | ||
| bbox_head=dict( | ||
| type='SharedDetHead', | ||
| num_classes=80,#len(classes), | ||
| in_channels=headfeat_channel, | ||
| feat_channels=headfeat_channel, | ||
| strides=[16,], | ||
| stacked_convs=0, | ||
| loss_bbox=dict( | ||
| type='IoULoss', | ||
| mode='square', | ||
| eps=1e-16, | ||
| reduction='sum', | ||
| loss_weight=5.0), | ||
| act_cfg=dict(type='ReLU'), | ||
| use_depthwise=True), | ||
| train_cfg=dict(assigner=dict(type='SimOTAAssigner', center_radius=2.5)), | ||
| test_cfg=dict(score_thr=0.01, nms=dict(type='nms', iou_threshold=0.65))) | ||
| #data_root = 'data/coco/' | ||
| #dataset_type = 'CocoDataset' | ||
| train_pipeline = [ | ||
| dict(type='Mosaic', img_scale=img_scale, pad_val=114.0, prob=0.5), | ||
| dict( | ||
| type='RandomAffine', | ||
| superposition=False, | ||
| scaling_ratio_range=(0.5, 1.5), | ||
| border=(-img_scale[0]//2, -img_scale[1]//2)), | ||
| dict( | ||
| type='FemtoDetMixUp', | ||
| superposition=False, | ||
| img_scale=img_scale, | ||
| ratio_range=(0.8, 1.2), | ||
| pad_val=114.0), | ||
| dict(type='YOLOXHSVRandomAug_EodVersion', hgain=0.015, sgain=0.7, vgain=0.4), | ||
| dict(type='RandomFlip', flip_ratio=0.5), | ||
| dict(type='Resize', img_scale=img_scale, keep_ratio=True), | ||
| dict( | ||
| type='Pad', | ||
| pad_to_square=True, | ||
| pad_val=dict(img=(114.0, 114.0, 114.0))), | ||
| dict(type='Normalize', **img_norm_cfg), | ||
| dict(type='FilterAnnotations', min_gt_bbox_wh=(1, 1), keep_empty=False), | ||
| dict(type='DefaultFormatBundle'), | ||
| dict(type='Collect', keys=['img', 'gt_bboxes', 'gt_labels']) | ||
| ] | ||
| train_dataset = dict( | ||
| type='MultiImageMixDataset', | ||
| dataset=dict( | ||
| type='CocoDataset', | ||
| # classes=classes, | ||
| ann_file=data_root+'annotations/instances_train2017.json', | ||
| img_prefix=data_root+'train2017', | ||
| pipeline=[ | ||
| dict(type='LoadImageFromFile'), | ||
| dict(type='LoadAnnotations', with_bbox=True) | ||
| ], | ||
| filter_empty_gt=False), | ||
| pipeline=train_pipeline) | ||
|
|
||
| test_pipeline = [ | ||
| dict(type='LoadImageFromFile'), | ||
| dict( | ||
| type='MultiScaleFlipAug', | ||
| img_scale=test_img_scale, | ||
| flip=False, | ||
| transforms=[ | ||
| dict(type='Resize', keep_ratio=True), | ||
| dict(type='RandomFlip'), | ||
| dict( | ||
| type='Pad', | ||
| pad_to_square=True, | ||
| pad_val=dict(img=(114.0, 114.0, 114.0))), | ||
| dict(type='Normalize', **img_norm_cfg), | ||
| dict(type='DefaultFormatBundle'), | ||
| dict(type='Collect', keys=['img']) | ||
| ]) | ||
| ] | ||
| data = dict( | ||
| samples_per_gpu=8, | ||
| workers_per_gpu=4, | ||
| persistent_workers=True, | ||
| train=train_dataset, | ||
| val=dict( | ||
| type='CocoDataset', | ||
| # classes=classes, | ||
| ann_file=data_root+'annotations/instances_val2017.json', | ||
| img_prefix=data_root+'val2017', | ||
| pipeline=test_pipeline), | ||
| test=dict( | ||
| type='CocoDataset', | ||
| # classes=classes, | ||
| ann_file=data_root+'annotations/instances_val2017.json', | ||
| img_prefix=data_root+'val2017', | ||
| pipeline=test_pipeline)) | ||
|
|
||
| interval = 10 | ||
| evaluation = dict( | ||
| save_best='auto', interval=10, dynamic_intervals=[(285, 1)], metric='bbox') |
174 changes: 174 additions & 0 deletions
174
configs/femtoDet_large_coco/femtodet_large_coco_1stage.py
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,174 @@ | ||
| #dataset | ||
| # classes = ('aeroplane', 'bicycle', 'bird', 'boat', 'bottle', 'bus', 'car', 'cat', 'chair', 'cow', 'diningtable', | ||
| # 'dog', 'horse', 'motorbike', 'person', 'pottedplant', 'sheep', 'sofa', 'train', 'tvmonitor') | ||
| data_root = './data/coco/' | ||
| #model | ||
| img_scale = (640, 640) | ||
| widen_factor=1.0 | ||
| num_last_epochs=0 | ||
| max_epochs=300 | ||
| test_img_scale=(416,416) | ||
| pretrain_weight='./weights/femto_large_imagenet1k.pth' | ||
| load_from = './work_dirs/femtodet_large_coco_0stage/epoch_300.pth' | ||
| resume_from = None | ||
| #more information | ||
| optimizer = dict( | ||
| type='SGD', | ||
| lr=0.01, | ||
| momentum=0.90, | ||
| weight_decay=0.0005, | ||
| nesterov=True, | ||
| paramwise_cfg=dict(norm_decay_mult=0.0, bias_decay_mult=0.0)) | ||
| optimizer_config = dict(grad_clip=None) | ||
| lr_config = dict( | ||
| policy='YOLOX', | ||
| warmup='exp', | ||
| by_epoch=False, | ||
| warmup_by_epoch=True, | ||
| warmup_ratio=1, | ||
| warmup_iters=5, | ||
| num_last_epochs=num_last_epochs, | ||
| min_lr_ratio=0.05) | ||
| runner = dict(type='EpochBasedRunner', max_epochs=max_epochs) | ||
| checkpoint_config = dict(interval=10) | ||
| log_config = dict(interval=50, hooks=[dict(type='TextLoggerHook')]) | ||
| custom_hooks = [ | ||
| dict(type='YOLOXModeSwitchHook', num_last_epochs=num_last_epochs, priority=48),# all_unfreeze_after_this_epoch=50), | ||
| dict(type='SyncNormHook', num_last_epochs=num_last_epochs, interval=10, priority=48), | ||
| dict( | ||
| type='ExpMomentumEMAHook', | ||
| resume_from=None, | ||
| momentum=0.0001, | ||
| priority=49) | ||
| ] | ||
| find_unused_parameters=True | ||
| dist_params = dict(backend='nccl') | ||
| log_level = 'INFO' | ||
| workflow = [('train', 1)] | ||
| opencv_num_threads = 0 | ||
| mp_start_method = 'fork' | ||
| default_channels=[32, 96, 320] | ||
| img_norm_cfg = dict(mean=[123.675, 116.28, 103.53], std=[58.82, 58.82, 58.82], to_rgb=True) | ||
| neck_in_chanels = [int(ch*widen_factor) for ch in default_channels] | ||
| headfeat_channel = 64 | ||
| model = dict( | ||
| type='FemtoDet', | ||
| input_size=img_scale, | ||
| random_size_range=(10, 20), | ||
| random_size_interval=10, | ||
| backbone=dict( | ||
| type='FemtoNet', | ||
| widen_factor=widen_factor, | ||
| diff_conv=True, | ||
| out_indices=(2, 4, 6), | ||
| act_cfg=dict(type='ReLU'), | ||
| init_cfg=dict( | ||
| type='Pretrained', | ||
| prefix='backbone', | ||
| checkpoint=pretrain_weight)), | ||
| neck=dict( | ||
| type='SharedNeck', | ||
| in_channels=neck_in_chanels, | ||
| out_channels=headfeat_channel, | ||
| fixed_size_idx=1, | ||
| add=True, | ||
| norm_cfg=dict(type='BN'), | ||
| act_cfg=dict(type='ReLU'), | ||
| num_outs=1), | ||
| bbox_head=dict( | ||
| type='SharedDetHead', | ||
| num_classes=80, #len(classes), | ||
| in_channels=headfeat_channel, | ||
| feat_channels=headfeat_channel, | ||
| strides=[16,], | ||
| stacked_convs=0, | ||
| loss_bbox=dict( | ||
| type='IoULoss', | ||
| mode='square', | ||
| eps=1e-16, | ||
| reduction='sum', | ||
| loss_weight=5.0), | ||
| act_cfg=dict(type='ReLU'), | ||
| use_depthwise=True), | ||
| train_cfg=dict(assigner=dict(type='SimOTAAssigner', center_radius=2.5)), | ||
| test_cfg=dict(score_thr=0.01, nms=dict(type='nms', iou_threshold=0.65))) | ||
| #data_root = 'data/coco/' | ||
| #dataset_type = 'CocoDataset' | ||
| train_pipeline = [ | ||
| dict(type='Mosaic', img_scale=img_scale, pad_val=114.0, prob=0.5), | ||
| dict( | ||
| type='RandomAffine', | ||
| superposition=False, | ||
| scaling_ratio_range=(0.5, 1.5), | ||
| border=(-img_scale[0]//2, -img_scale[1]//2)), | ||
| # dict( | ||
| # type='FemtoDetMixUp', | ||
| # superposition=False, | ||
| # img_scale=img_scale, | ||
| # ratio_range=(0.8, 1.2), | ||
| # pad_val=114.0), | ||
| dict(type='YOLOXHSVRandomAug_EodVersion', hgain=0.015, sgain=0.7, vgain=0.4), | ||
| dict(type='RandomFlip', flip_ratio=0.5), | ||
| dict(type='Resize', img_scale=img_scale, keep_ratio=True), | ||
| dict( | ||
| type='Pad', | ||
| pad_to_square=True, | ||
| pad_val=dict(img=(114.0, 114.0, 114.0))), | ||
| dict(type='Normalize', **img_norm_cfg), | ||
| dict(type='FilterAnnotations', min_gt_bbox_wh=(1, 1), keep_empty=False), | ||
| dict(type='DefaultFormatBundle'), | ||
| dict(type='Collect', keys=['img', 'gt_bboxes', 'gt_labels']) | ||
| ] | ||
| train_dataset = dict( | ||
| type='MultiImageMixDataset', | ||
| dataset=dict( | ||
| type='CocoDataset', | ||
| # classes=classes, | ||
| ann_file=data_root+'annotations/instances_train2017.json', | ||
| img_prefix=data_root+'train2017', | ||
| pipeline=[ | ||
| dict(type='LoadImageFromFile'), | ||
| dict(type='LoadAnnotations', with_bbox=True) | ||
| ], | ||
| filter_empty_gt=False), | ||
| pipeline=train_pipeline) | ||
|
|
||
| test_pipeline = [ | ||
| dict(type='LoadImageFromFile'), | ||
| dict( | ||
| type='MultiScaleFlipAug', | ||
| img_scale=test_img_scale, | ||
| flip=False, | ||
| transforms=[ | ||
| dict(type='Resize', keep_ratio=True), | ||
| dict(type='RandomFlip'), | ||
| dict( | ||
| type='Pad', | ||
| pad_to_square=True, | ||
| pad_val=dict(img=(114.0, 114.0, 114.0))), | ||
| dict(type='Normalize', **img_norm_cfg), | ||
| dict(type='DefaultFormatBundle'), | ||
| dict(type='Collect', keys=['img']) | ||
| ]) | ||
| ] | ||
| data = dict( | ||
| samples_per_gpu=8, | ||
| workers_per_gpu=4, | ||
| persistent_workers=True, | ||
| train=train_dataset, | ||
| val=dict( | ||
| type='CocoDataset', | ||
| # classes=classes, | ||
| ann_file=data_root+'annotations/instances_val2017.json', | ||
| img_prefix=data_root+'val2017', | ||
| pipeline=test_pipeline), | ||
| test=dict( | ||
| type='CocoDataset', | ||
| # classes=classes, | ||
| ann_file=data_root+'annotations/instances_val2017.json', | ||
| img_prefix=data_root+'val2017', | ||
| pipeline=test_pipeline)) | ||
|
|
||
| interval = 10 | ||
| evaluation = dict( | ||
| save_best='auto', interval=10, dynamic_intervals=[(285, 1)], metric='bbox') |
Oops, something went wrong.