Fix roibugs

README.md

@@ -55,6 +55,8 @@ FAIR Dataset: [fair.md](docs/fair.md)
 
 SSDD/SSDD+: [ssdd.md](docs/ssdd.md)
 
+HRSC2016: [hrsc2016.md](docs/hrsc2016.md)
+
 You can also build your own dataset by convert your datas to DOTA format.
 ### Config
 JDet defines the used model, dataset and training/testing method by `config-file`, please check the [config.md](docs/config.md) to learn how it works.

configs/faster_rcnn_RoITrans_r50_fpn_1x_dota.py

@@ -144,7 +144,7 @@
     workers_per_gpu=2,
     train=dict(
         type=dataset_type,
-        dataset_dir='/home/cxjyxx_me/workspace/JAD/datasets/processed_DOTA/trainval_1024_200_1.0',
+        dataset_dir='/home/flowey/dataset/processed_DOTA/trainval_1024_200_1.0/',
         version='1',
         filter_min_size=32,
         transforms=[
@@ -163,10 +163,11 @@
                 to_bgr=True),
         ],
         batch_size=2,
+        shuffle=True,
         ),
     val=dict(
         type=dataset_type,
-        dataset_dir='/home/cxjyxx_me/workspace/JAD/datasets/processed_DOTA/trainval_1024_200_1.0',
+        dataset_dir='/home/flowey/dataset/processed_DOTA/trainval_1024_200_1.0/',
         version='1',
         filter_min_size=32,
         transforms=[
@@ -182,7 +183,7 @@
         ),
     test=dict(
         type="ImageDataset",        
-        images_dir='/mnt/disk/lxl/dataset/DOTA_1024/test_split/images/',
+        images_dir='/home/flowey/dataset/processed_DOTA/test_1024_200_1.0/images',
         transforms=[
             dict(
                 type = "Pad",

configs/preprocess/hrsc2016_preprocess_config.py

@@ -0,0 +1,23 @@
+type='HRSC2016'
+source_dataset_path='/home/flowey/dataset/HRSC2016'
+
+tasks=[
+    dict(
+        label='train',
+        config=dict(
+            images_path=source_dataset_path+'/Train/AllImages',
+            xml_path=source_dataset_path+'/Train/Annotations',
+            imageset_file=source_dataset_path+'/Train/train.txt',
+            out_annotation_file=source_dataset_path+'/Train/labels.pkl',
+        )
+    ),
+    dict(
+        label='test',
+        config=dict(
+            images_path=source_dataset_path+'/Test/AllImages',
+            xml_path=source_dataset_path+'/Test/Annotations',
+            imageset_file=source_dataset_path+'/Test/test.txt',
+            out_annotation_file=source_dataset_path+'/Test/labels.pkl',
+        )
+    )
+]

configs/s2anet/s2anet_r50_fpn_3x_hrsc2016.py

@@ -0,0 +1,183 @@
+# model settings
+model = dict(
+    type='S2ANet',
+    backbone=dict(
+        type='Resnet50',
+        frozen_stages=1,
+        return_stages=["layer1","layer2","layer3","layer4"],
+        pretrained= True),
+    neck=dict(
+        type='FPN',
+        in_channels=[256, 512, 1024, 2048],
+        out_channels=256,
+        start_level=1,
+        add_extra_convs="on_input",
+        num_outs=5),
+    bbox_head=dict(
+        type='S2ANetHead',
+        num_classes=2,
+        in_channels=256,
+        feat_channels=256,
+        stacked_convs=2,
+        with_orconv=True,
+        anchor_ratios=[1.0],
+        anchor_strides=[8, 16, 32, 64, 128],
+        anchor_scales=[4],
+        target_means=[.0, .0, .0, .0, .0],
+        target_stds=[1.0, 1.0, 1.0, 1.0, 1.0],
+        loss_fam_cls=dict(
+            type='FocalLoss',
+            use_sigmoid=True,
+            gamma=2.0,
+            alpha=0.25,
+            loss_weight=1.0),
+        loss_fam_bbox=dict(
+            type='SmoothL1Loss', beta=1.0 / 9.0, loss_weight=1.0),
+        loss_odm_cls=dict(
+            type='FocalLoss',
+            use_sigmoid=True,
+            gamma=2.0,
+            alpha=0.25,
+            loss_weight=1.0),
+        loss_odm_bbox=dict(
+            type='SmoothL1Loss', beta=1.0 / 9.0, loss_weight=1.0),
+        test_cfg=dict(
+            nms_pre=2000,
+            min_bbox_size=0,
+            score_thr=0.05,
+            nms=dict(type='nms_rotated', iou_thr=0.1),
+            max_per_img=2000),
+        train_cfg=dict(
+            fam_cfg=dict(
+                assigner=dict(
+                    type='MaxIoUAssigner',
+                    pos_iou_thr=0.5,
+                    neg_iou_thr=0.4,
+                    min_pos_iou=0,
+                    ignore_iof_thr=-1,
+                    iou_calculator=dict(type='BboxOverlaps2D_rotated')),
+                bbox_coder=dict(type='DeltaXYWHABBoxCoder',
+                                target_means=(0., 0., 0., 0., 0.),
+                                target_stds=(1., 1., 1., 1., 1.),
+                                clip_border=True),
+                allowed_border=-1,
+                pos_weight=-1,
+                debug=False),
+            odm_cfg=dict(
+                assigner=dict(
+                    type='MaxIoUAssigner',
+                    pos_iou_thr=0.5,
+                    neg_iou_thr=0.4,
+                    min_pos_iou=0,
+                    ignore_iof_thr=-1,
+                    iou_calculator=dict(type='BboxOverlaps2D_rotated')),
+                bbox_coder=dict(type='DeltaXYWHABBoxCoder',
+                                target_means=(0., 0., 0., 0., 0.),
+                                target_stds=(1., 1., 1., 1., 1.),
+                                clip_border=True),
+                allowed_border=-1,
+                pos_weight=-1,
+                debug=False))
+        )
+    )
+base_data_root = '/home/flowey/dataset/HRSC2016'
+dataset = dict(
+    train=dict(
+        type="HRSC2016Dataset",
+        images_dir = base_data_root + '/Train/AllImages',
+        annotations_file = base_data_root + '/Train/labels.pkl',
+        transforms=[
+            dict(
+                type="RotatedResize",
+                min_size=512,
+                max_size=800
+            ),
+            dict(type='RotatedRandomFlip', prob=0.5),
+            dict(
+                type = "Pad",
+                size_divisor=32),
+            dict(
+                type = "Normalize",
+                mean =  [123.675, 116.28, 103.53],
+                std = [58.395, 57.12, 57.375],
+                to_bgr=False,)
+
+        ],
+        batch_size=2,
+        num_workers=4,
+        shuffle=True,
+        filter_empty_gt=False
+    ),
+    val=dict(
+        type="HRSC2016Dataset",
+        images_dir = base_data_root + '/Test/AllImages',
+        annotations_file = base_data_root + '/Test/labels.pkl',
+        use_07_metric=False,
+        transforms=[
+            dict(
+                type="RotatedResize",
+                min_size=512,
+                max_size=800
+            ),
+            dict(
+                type = "Pad",
+                size_divisor=32),
+            dict(
+                type = "Normalize",
+                mean =  [123.675, 116.28, 103.53],
+                std = [58.395, 57.12, 57.375],
+                to_bgr=False),
+        ],
+        batch_size=2,
+        num_workers=4,
+        shuffle=False
+    ),
+    test=dict(
+        type="ImageDataset",
+        dataset_type="HRSC2016",
+        images_dir = base_data_root + '/Test/AllImages',
+        transforms=[
+            dict(
+                type="RotatedResize",
+                min_size=512,
+                max_size=800
+            ),
+            dict(
+                type = "Pad",
+                size_divisor=32),
+            dict(
+                type = "Normalize",
+                mean =  [123.675, 116.28, 103.53],
+                std = [58.395, 57.12, 57.375],
+                to_bgr=False,),
+        ],
+        num_workers=4,
+        batch_size=2,
+    )
+)
+
+optimizer = dict(
+    type='SGD', 
+    lr=0.01/4., #0.0,#0.01*(1/8.), 
+    momentum=0.9, 
+    weight_decay=0.0001,
+    grad_clip=dict(
+        max_norm=35, 
+        norm_type=2))
+
+scheduler = dict(
+    type='StepLR',
+    warmup='linear',
+    warmup_iters=500,
+    warmup_ratio=1.0 / 3,
+    milestones=[24, 33])
+
+
+logger = dict(
+    type="RunLogger")
+
+# when we the trained model from cshuan, image is rgb
+max_epoch = 36
+eval_interval = 1
+checkpoint_interval = 1
+log_interval = 50

configs/san10_pairwise.py

@@ -0,0 +1,98 @@
+model = dict(
+    type='SAN',
+    sa_type=0,
+    layers=[2, 1, 2, 4, 1],
+    kernels=[3, 7, 7, 7, 7],
+    num_classes=1000,
+    loss=dict(
+        type='SAMSmoothLoss',
+        eps=0.1
+    ),
+    loss_prepare=False
+)
+
+# dataset settings
+dataset_type = 'ILSVRCDataset'
+dataset = dict(
+    imgs_per_gpu=2,
+    workers_per_gpu=4,
+    train=dict(
+        type=dataset_type,
+        images_dir='/home/flowey/dataset/ILSVRC2012/train/',
+        transforms=[
+            dict(type = "Resize",           ## TODO: implement RandomRotatedCrop
+                 min_size = 224,
+                 max_size = 224,
+            ),
+            dict(
+                type = "RotatedRandomFlip",
+                prob = 0.5,
+                direction="horizontal",
+            ),
+            dict(
+                type = "Normalize",         ## unknown normalize
+                mean =  [123.675, 116.28, 103.53],
+                std = [58.395, 57.12, 57.375],
+                to_bgr=False),
+        ],
+        batch_size=2,
+        ),
+    val=dict(
+        type=dataset_type,
+        batch_size=100,
+        images_dir='/home/flowey/dataset/ILSVRC2012/val/',
+        transforms=[
+            dict(type = "Resize",
+                 min_size = 256,
+                 max_size = None,
+            ),
+            dict(type = "CenterCropJt",
+                 size = 224,
+            ),
+            dict(
+                type = "Normalize",
+                mean =  [123.675, 116.28, 103.53],
+                std = [58.395, 57.12, 57.375],
+                to_bgr=False),
+        ],
+        ),
+    test=dict(
+        type="ImageDataset",        
+        images_dir='/mnt/disk/lxl/dataset/DOTA_1024/test_split/images/',
+        transforms=[
+            dict(type = "Resize",
+                 min_size = 224,
+                 max_size = 224,
+            ),
+            dict(
+                type = "Normalize",
+                mean =  [123.675, 116.28, 103.53],
+                std = [58.395, 57.12, 57.375],
+                to_bgr=False),
+        ],
+    )
+)
+# optimizer
+optimizer = dict(
+    type='SGD', 
+    lr=0.01, 
+    momentum=0.9, 
+    weight_decay=0.0001,
+    grad_clip=dict(         ##grad_clip: not sure
+        max_norm=35, 
+        norm_type=2))
+
+# learning policy
+scheduler = dict(
+    type='CosineAnnealingLR',
+    warmup='linear',        ##warmup: not sure
+    warmup_iters=500,
+    warmup_ratio=1.0 / 3,
+    max_steps=100)
+
+logger = dict(
+    type="RunLogger")
+max_epoch = 100
+eval_interval = 25
+checkpoint_interval = 10
+log_interval = 20