update codes

configs/grounding_dino/dod/grounding_dino_swin-b_pretrain_zeroshot_concat_dod.py

+_base_ = 'grounding_dino_swin-t_pretrain_zeroshot_concat_dod.py'
+
+model = dict(
+    type='GroundingDINO',
+    backbone=dict(
+        pretrain_img_size=384,
+        embed_dims=128,
+        depths=[2, 2, 18, 2],
+        num_heads=[4, 8, 16, 32],
+        window_size=12,
+        drop_path_rate=0.3,
+        patch_norm=True),
+    neck=dict(in_channels=[256, 512, 1024]),
+)

configs/grounding_dino/dod/grounding_dino_swin-b_pretrain_zeroshot_parallel_dod.py

+_base_ = 'grounding_dino_swin-b_pretrain_zeroshot_concat_dod.py'
+
+model = dict(test_cfg=dict(chunked_size=1))

configs/grounding_dino/dod/grounding_dino_swin-t_pretrain_zeroshot_concat_dod.py

+_base_ = '../grounding_dino_swin-t_pretrain_obj365_goldg_cap4m.py'
+
+data_root = 'data/d3/'
+
+test_pipeline = [
+    dict(
+        type='LoadImageFromFile', backend_args=None,
+        imdecode_backend='pillow'),
+    dict(
+        type='FixScaleResize',
+        scale=(800, 1333),
+        keep_ratio=True,
+        backend='pillow'),
+    dict(type='LoadAnnotations', with_bbox=True),
+    dict(
+        type='PackDetInputs',
+        meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape',
+                   'scale_factor', 'text', 'custom_entities', 'sent_ids'))
+]
+
+# -------------------------------------------------#
+val_dataset_full = dict(
+    type='DODDataset',
+    data_root=data_root,
+    ann_file='d3_json/d3_full_annotations.json',
+    data_prefix=dict(img='d3_images/', anno='d3_pkl'),
+    pipeline=test_pipeline,
+    test_mode=True,
+    backend_args=None,
+    return_classes=True)
+
+val_evaluator_full = dict(
+    type='DODCocoMetric',
+    ann_file=data_root + 'd3_json/d3_full_annotations.json')
+
+# -------------------------------------------------#
+val_dataset_pres = dict(
+    type='DODDataset',
+    data_root=data_root,
+    ann_file='d3_json/d3_pres_annotations.json',
+    data_prefix=dict(img='d3_images/', anno='d3_pkl'),
+    pipeline=test_pipeline,
+    test_mode=True,
+    backend_args=None,
+    return_classes=True)
+val_evaluator_pres = dict(
+    type='DODCocoMetric',
+    ann_file=data_root + 'd3_json/d3_pres_annotations.json')
+
+# -------------------------------------------------#
+val_dataset_abs = dict(
+    type='DODDataset',
+    data_root=data_root,
+    ann_file='d3_json/d3_abs_annotations.json',
+    data_prefix=dict(img='d3_images/', anno='d3_pkl'),
+    pipeline=test_pipeline,
+    test_mode=True,
+    backend_args=None,
+    return_classes=True)
+val_evaluator_abs = dict(
+    type='DODCocoMetric',
+    ann_file=data_root + 'd3_json/d3_abs_annotations.json')
+
+# -------------------------------------------------#
+datasets = [val_dataset_full, val_dataset_pres, val_dataset_abs]
+dataset_prefixes = ['FULL', 'PRES', 'ABS']
+metrics = [val_evaluator_full, val_evaluator_pres, val_evaluator_abs]
+
+val_dataloader = dict(
+    dataset=dict(_delete_=True, type='ConcatDataset', datasets=datasets))
+test_dataloader = val_dataloader
+
+val_evaluator = dict(
+    _delete_=True,
+    type='MultiDatasetsEvaluator',
+    metrics=metrics,
+    dataset_prefixes=dataset_prefixes)
+test_evaluator = val_evaluator

configs/grounding_dino/dod/grounding_dino_swin-t_pretrain_zeroshot_parallel_dod.py

+_base_ = 'grounding_dino_swin-t_pretrain_zeroshot_concat_dod.py'
+
+model = dict(test_cfg=dict(chunked_size=1))

configs/grounding_dino/flickr30k/grounding_dino_swin-t-pretrain_zeroshot_flickr30k.py

+_base_ = '../grounding_dino_swin-t_pretrain_obj365_goldg_cap4m.py'
+
+dataset_type = 'Flickr30kDataset'
+data_root = 'data/flickr30k_entities/'
+
+test_pipeline = [
+    dict(
+        type='LoadImageFromFile', backend_args=None,
+        imdecode_backend='pillow'),
+    dict(
+        type='FixScaleResize',
+        scale=(800, 1333),
+        keep_ratio=True,
+        backend='pillow'),
+    dict(type='LoadAnnotations', with_bbox=True),
+    dict(
+        type='PackDetInputs',
+        meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape',
+                   'scale_factor', 'text', 'custom_entities',
+                   'tokens_positive', 'phrase_ids', 'phrases'))
+]
+
+dataset_Flickr30k_val = dict(
+    type=dataset_type,
+    data_root=data_root,
+    ann_file='final_flickr_separateGT_val.json',
+    data_prefix=dict(img='flickr30k_images/'),
+    pipeline=test_pipeline,
+)
+
+dataset_Flickr30k_test = dict(
+    type=dataset_type,
+    data_root=data_root,
+    ann_file='final_flickr_separateGT_test.json',
+    data_prefix=dict(img='flickr30k_images/'),
+    pipeline=test_pipeline,
+)
+
+val_evaluator_Flickr30k = dict(type='Flickr30kMetric')
+
+test_evaluator_Flickr30k = dict(type='Flickr30kMetric')
+
+# ----------Config---------- #
+dataset_prefixes = ['Flickr30kVal', 'Flickr30kTest']
+datasets = [dataset_Flickr30k_val, dataset_Flickr30k_test]
+metrics = [val_evaluator_Flickr30k, test_evaluator_Flickr30k]
+
+val_dataloader = dict(
+    dataset=dict(_delete_=True, type='ConcatDataset', datasets=datasets))
+test_dataloader = val_dataloader
+
+val_evaluator = dict(
+    _delete_=True,
+    type='MultiDatasetsEvaluator',
+    metrics=metrics,
+    dataset_prefixes=dataset_prefixes)
+test_evaluator = val_evaluator

configs/grounding_dino/grounding_dino_swin-b_finetune_16xb2_1x_coco.py

+_base_ = [
+    './grounding_dino_swin-t_finetune_16xb2_1x_coco.py',
+]
+
+load_from = 'https://download.openmmlab.com/mmdetection/v3.0/grounding_dino/groundingdino_swinb_cogcoor_mmdet-55949c9c.pth'  # noqa
+model = dict(
+    type='GroundingDINO',
+    backbone=dict(
+        pretrain_img_size=384,
+        embed_dims=128,
+        depths=[2, 2, 18, 2],
+        num_heads=[4, 8, 16, 32],
+        window_size=12,
+        drop_path_rate=0.3,
+        patch_norm=True),
+    neck=dict(in_channels=[256, 512, 1024]),
+)

configs/grounding_dino/grounding_dino_swin-b_pretrain_mixeddata.py

+_base_ = [
+    './grounding_dino_swin-t_pretrain_obj365_goldg_cap4m.py',
+]
+
+model = dict(
+    type='GroundingDINO',
+    backbone=dict(
+        pretrain_img_size=384,
+        embed_dims=128,
+        depths=[2, 2, 18, 2],
+        num_heads=[4, 8, 16, 32],
+        window_size=12,
+        drop_path_rate=0.3,
+        patch_norm=True),
+    neck=dict(in_channels=[256, 512, 1024]),
+)

configs/grounding_dino/grounding_dino_swin-t_finetune_8xb2_20e_cat.py

+_base_ = 'grounding_dino_swin-t_finetune_16xb2_1x_coco.py'
+
+data_root = 'data/cat/'
+class_name = ('cat', )
+num_classes = len(class_name)
+metainfo = dict(classes=class_name, palette=[(220, 20, 60)])
+
+model = dict(bbox_head=dict(num_classes=num_classes))
+
+train_dataloader = dict(
+    dataset=dict(
+        data_root=data_root,
+        metainfo=metainfo,
+        ann_file='annotations/trainval.json',
+        data_prefix=dict(img='images/')))
+
+val_dataloader = dict(
+    dataset=dict(
+        metainfo=metainfo,
+        data_root=data_root,
+        ann_file='annotations/test.json',
+        data_prefix=dict(img='images/')))
+
+test_dataloader = val_dataloader
+
+val_evaluator = dict(ann_file=data_root + 'annotations/test.json')
+test_evaluator = val_evaluator
+
+max_epoch = 20
+
+default_hooks = dict(
+    checkpoint=dict(interval=1, max_keep_ckpts=1, save_best='auto'),
+    logger=dict(type='LoggerHook', interval=5))
+train_cfg = dict(max_epochs=max_epoch, val_interval=1)
+
+param_scheduler = [
+    dict(type='LinearLR', start_factor=0.001, by_epoch=False, begin=0, end=30),
+    dict(
+        type='MultiStepLR',
+        begin=0,
+        end=max_epoch,
+        by_epoch=True,
+        milestones=[15],
+        gamma=0.1)
+]
+
+optim_wrapper = dict(
+    optimizer=dict(lr=0.00005),
+    paramwise_cfg=dict(
+        custom_keys={
+            'absolute_pos_embed': dict(decay_mult=0.),
+            'backbone': dict(lr_mult=0.1),
+            'language_model': dict(lr_mult=0),
+        }))
+
+auto_scale_lr = dict(base_batch_size=16)