paa_dcnv2_X_101_64x4d_FPN_2x.yaml

MODEL:
  META_ARCHITECTURE: "GeneralizedRCNN"
  WEIGHT: "catalog://ImageNetPretrained/FAIR/20171220/X-101-64x4d"
  RPN_ONLY: True
  PAA_ON: True
  BACKBONE:
    CONV_BODY: "R-101-FPN-RETINANET"
  RESNETS:
    STRIDE_IN_1X1: False
    BACKBONE_OUT_CHANNELS: 256
    NUM_GROUPS: 64
    WIDTH_PER_GROUP: 4
    STAGE_WITH_DCN: (False, False, True, True)
    WITH_MODULATED_DCN: True
    DEFORMABLE_GROUPS: 1
  RETINANET:
    USE_C5: False
  PAA:
    ANCHOR_SIZES: (64, 128, 256, 512, 1024) # 8S
    ASPECT_RATIOS: (1.0,)
    SCALES_PER_OCTAVE: 1
    USE_DCN_IN_TOWER: True
    TOPK: 9 # topk for selecting candidate positive samples from each level
    IOU_THRESHOLD: 0.1
    REG_LOSS_WEIGHT: 1.3
    USE_IOU_PRED: True
    IOU_LOSS_WEIGHT: 0.5
    INFERENCE_SCORE_VOTING: True
DATASETS:
  TRAIN: ("coco_2017_train",)
  TEST: ("coco_2017_val",)
INPUT:
  MIN_SIZE_RANGE_TRAIN: (640, 800)
  MAX_SIZE_TRAIN: 1333
  MIN_SIZE_TEST: 800
  MAX_SIZE_TEST: 1333
DATALOADER:
  SIZE_DIVISIBILITY: 32
SOLVER:
  BASE_LR: 0.01
  WEIGHT_DECAY: 0.0001
  STEPS: (120000, 160000)
  MAX_ITER: 180000
  IMS_PER_BATCH: 16
  WARMUP_METHOD: "constant"
TEST:
  BBOX_AUG:
    ENABLED: False
    VOTE: True
    VOTE_TH: 0.66
    MERGE_TYPE: "soft-vote"
    H_FLIP: True
    SCALES: (400, 500, 600, 640, 700, 900, 1000, 1100, 1200, 1300, 1400, 1800)
    SCALE_RANGES: [[96, 10000], [96, 10000], [64, 10000], [64, 10000], [64, 10000], [0, 10000], [0, 10000], [0, 256], [0, 256], [0, 192], [0, 192], [0, 96]]
    MAX_SIZE: 3000
    SCALE_H_FLIP: True