add dyhead configs & weights

dyhead/atss_r50_config.yaml

@@ -0,0 +1,28 @@
+MODEL:
+  META_ARCHITECTURE: "ATSS"
+  WEIGHTS: "detectron2://ImageNetPretrained/MSRA/R-50.pkl"
+  BACKBONE:
+    NAME: "build_retinanet_resnet_fpn_dyhead_backbone"
+  RESNETS:
+    DEPTH: 50
+    OUT_FEATURES: ["res3", "res4", "res5"]
+  FPN:
+    IN_FEATURES: ["res3", "res4", "res5"]
+  DYHEAD:
+    NUM_CONVS: 6
+    CHANNELS: 256
+  ANCHOR_GENERATOR:
+    SIZES: !!python/object/apply:eval ["[[x*2,] for x in [32, 64, 128, 256, 512 ]]"]
+    ASPECT_RATIOS: [1.0,]
+    OFFSET: 0.5
+  ATSS:
+    NUM_CONVS: 0
+DATASETS:
+  TRAIN: ("coco_2017_train",)
+  TEST: ("coco_2017_val",)
+SOLVER:
+  IMS_PER_BATCH: 16
+  BASE_LR: 0.01
+  STEPS: (60000, 80000)
+  MAX_ITER: 90000
+VERSION: 2

dyhead/atss_r50_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c2eee9e73748ee6152ba5edce00a8b0579b808ae340d3040f540bc8b7ca8ecf3
+size 155972489

dyhead/atss_swin_tiny_config.yaml

@@ -0,0 +1,34 @@
+MODEL:
+  META_ARCHITECTURE: "ATSS"
+  WEIGHTS: "swin_tiny_patch4_window7_224_d2.pth"
+  PIXEL_MEAN: [123.675, 116.28, 103.53]
+  PIXEL_STD: [58.395, 57.12, 57.375]
+  BACKBONE:
+    NAME: "build_retinanet_swin_fpn_dyhead_backbone"
+  SWINT:
+    OUT_FEATURES: ["stage3", "stage4", "stage5"]
+  FPN:
+    IN_FEATURES: ["stage3", "stage4", "stage5"]
+  DYHEAD:
+    NUM_CONVS: 6
+    CHANNELS: 256
+  ANCHOR_GENERATOR:
+    SIZES: !!python/object/apply:eval ["[[x*2,] for x in [32, 64, 128, 256, 512 ]]"]
+    ASPECT_RATIOS: [1.0,]
+    OFFSET: 0.5
+  ATSS:
+    NUM_CONVS: 0
+DATASETS:
+  TRAIN: ("coco_2017_train",)
+  TEST: ("coco_2017_val",)
+SOLVER:
+  IMS_PER_BATCH: 32
+  BASE_LR: 0.0001
+  WEIGHT_DECAY: 0.05
+  STEPS: (60000, 80000)
+  MAX_ITER: 90000
+  OPTIMIZER: "ADAMW"
+INPUT:
+  FORMAT: "RGB"
+  MIN_SIZE_TRAIN: (640, 672, 704, 736, 768, 800)
+VERSION: 2

dyhead/atss_swin_tiny_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0c398fa3a779b4825e7d043d6451eca237db25386be509165654f0819dee8a26
+size 169728933

dyhead/faster_rcnn_r50_config.yaml

@@ -0,0 +1,45 @@
+MODEL:
+  META_ARCHITECTURE: "GeneralizedRCNN"
+  WEIGHTS: "detectron2://ImageNetPretrained/MSRA/R-50.pkl"
+  BACKBONE:
+    NAME: "build_resnet_fpn_dyhead_backbone"
+  RESNETS:
+    DEPTH: 50
+    OUT_FEATURES: ["res2", "res3", "res4", "res5"]
+  FPN:
+    IN_FEATURES: ["res2", "res3", "res4", "res5"]
+  DYHEAD:
+    NUM_CONVS: 6
+    CHANNELS: 256
+  ANCHOR_GENERATOR:
+    SIZES: [[32], [64], [128], [256], [512]]  # One size for each in feature map
+    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]  # Three aspect ratios (same for all in feature maps)
+  RPN:
+    IN_FEATURES: ["p2", "p3", "p4", "p5", "p6"]
+    PRE_NMS_TOPK_TRAIN: 2000  # Per FPN level
+    PRE_NMS_TOPK_TEST: 1000  # Per FPN level
+    # Detectron1 uses 2000 proposals per-batch,
+    # (See "modeling/rpn/rpn_outputs.py" for details of this legacy issue)
+    # which is approximately 1000 proposals per-image since the default batch size for FPN is 2.
+    POST_NMS_TOPK_TRAIN: 1000
+    POST_NMS_TOPK_TEST: 1000
+  ROI_HEADS:
+    NAME: "StandardROIHeads"
+    IN_FEATURES: ["p2", "p3", "p4", "p5"]
+  ROI_BOX_HEAD:
+    NAME: "FastRCNNConvFCHead"
+    NUM_FC: 2
+    POOLER_RESOLUTION: 7
+  ROI_MASK_HEAD:
+    NAME: "MaskRCNNConvUpsampleHead"
+    NUM_CONV: 4
+    POOLER_RESOLUTION: 14
+DATASETS:
+  TRAIN: ("coco_2017_train",)
+  TEST: ("coco_2017_val",)
+SOLVER:
+  IMS_PER_BATCH: 16
+  BASE_LR: 0.02
+  STEPS: (60000, 80000)
+  MAX_ITER: 90000
+VERSION: 2

dyhead/faster_rcnn_r50_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:35d7229f84f96c81cddb8d0c7b9a706939b0dd987e7733f8cd0ee9c75c560c82
+size 213392451

dyhead/retinanet_r50_config.yaml

@@ -0,0 +1,28 @@
+MODEL:
+  META_ARCHITECTURE: "RetinaNet"
+  WEIGHTS: "detectron2://ImageNetPretrained/MSRA/R-50.pkl"
+  BACKBONE:
+    NAME: "build_retinanet_resnet_fpn_dyhead_backbone"
+  RESNETS:
+    DEPTH: 50
+    OUT_FEATURES: ["res3", "res4", "res5"]
+  FPN:
+    IN_FEATURES: ["res3", "res4", "res5"]
+  DYHEAD:
+    NUM_CONVS: 6
+    CHANNELS: 256
+  ANCHOR_GENERATOR:
+    SIZES: !!python/object/apply:eval ["[[x, x * 2**(1.0/3), x * 2**(2.0/3) ] for x in [32, 64, 128, 256, 512 ]]"]
+  RETINANET:
+    IOU_THRESHOLDS: [0.4, 0.5]
+    IOU_LABELS: [0, -1, 1]
+    SMOOTH_L1_LOSS_BETA: 0.0
+DATASETS:
+  TRAIN: ("coco_2017_train",)
+  TEST: ("coco_2017_val",)
+SOLVER:
+  IMS_PER_BATCH: 16
+  BASE_LR: 0.01
+  STEPS: (60000, 80000)
+  MAX_ITER: 90000
+VERSION: 2

dyhead/retinanet_r50_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:798b595d6e45f826474eb97b2fe56cebf46b33923c24b03d92553efb3399b694
+size 181741239