_BASE_: ../Base-AVIS.yaml
MODEL:
  WEIGHTS: "./pre_models/SwinL_IN.pkl"
  BACKBONE:
    NAME: "D2SwinTransformer"
  SWIN:
    EMBED_DIM: 192
    DEPTHS: [2, 2, 18, 2]
    NUM_HEADS: [6, 12, 24, 48]
    WINDOW_SIZE: 12
    APE: False
    DROP_PATH_RATE: 0.3
    PATCH_NORM: True
    PRETRAIN_IMG_SIZE: 384
  PIXEL_MEAN: [123.675, 116.280, 103.530]
  PIXEL_STD: [58.395, 57.120, 57.375]
  META_ARCHITECTURE: "AVISM"
  MASK_ON: True
  SEM_SEG_HEAD:
    NAME: "MaskFormerHead"
    IGNORE_VALUE: 255
    NUM_CLASSES: 26
    LOSS_WEIGHT: 1.0
    CONVS_DIM: 256
    MASK_DIM: 256
    NORM: "GN"
    # pixel decoder
    PIXEL_DECODER_NAME: "MSDeformAttnPixelDecoder"
    IN_FEATURES: ["res2", "res3", "res4", "res5"]
    DEFORMABLE_TRANSFORMER_ENCODER_IN_FEATURES: ["res3", "res4", "res5"]
    COMMON_STRIDE: 4
    TRANSFORMER_ENC_LAYERS: 6
  MASK_FORMER:
    TRANSFORMER_DECODER_NAME: "AVISMMultiScaleMaskedTransformerDecoder"
    TRANSFORMER_IN_FEATURE: "multi_scale_pixel_decoder"
    DEEP_SUPERVISION: True
    NO_OBJECT_WEIGHT: 0.1
    CLASS_WEIGHT: 2.0
    MASK_WEIGHT: 5.0
    DICE_WEIGHT: 5.0
    HIDDEN_DIM: 256
    NUM_OBJECT_QUERIES: 200
    NHEADS: 8
    DROPOUT: 0.0
    DIM_FEEDFORWARD: 2048
    ENC_LAYERS: 0
    PRE_NORM: False
    ENFORCE_INPUT_PROJ: False
    SIZE_DIVISIBILITY: 32
    DEC_LAYERS: 10  # 9 decoder layers, add one for the loss on learnable query
    TRAIN_NUM_POINTS: 12544
    OVERSAMPLE_RATIO: 3.0
    IMPORTANCE_SAMPLE_RATIO: 0.75
    TEST:
      SEMANTIC_ON: False
      INSTANCE_ON: True
      PANOPTIC_ON: False
      OVERLAP_THRESHOLD: 0.8
      OBJECT_MASK_THRESHOLD: 0.8
  AVISM:
    ENC_WINDOW_SIZE: 6
    SIM_WEIGHT: 0.5
SOLVER:
  WARMUP_FACTOR: 1.0
  WARMUP_ITERS: 10
  WEIGHT_DECAY: 0.05
  OPTIMIZER: "ADAMW"
  BACKBONE_MULTIPLIER: 0.1
  CLIP_GRADIENTS:
    ENABLED: True
    CLIP_TYPE: "full_model"
    CLIP_VALUE: 0.01
    NORM_TYPE: 2.0
INPUT:
  SAMPLING_FRAME_NUM: 5
  SAMPLING_FRAME_RANGE: 20
  SAMPLING_FRAME_SHUFFLE: False
  # MIN_SIZE_TRAIN_SAMPLING : ["range", "choice", "range_by_clip", "choice_by_clip"]
  MIN_SIZE_TRAIN_SAMPLING: "choice_by_clip"
  # RANDOM_FLIP : ["none", "horizontal", "flip_by_clip"]. "horizontal" is set by default.
  RANDOM_FLIP: "flip_by_clip"
  AUGMENTATIONS: []
  MIN_SIZE_TRAIN: (288, 320, 352, 384, 416, 448, 480, 512)
  MAX_SIZE_TRAIN: 768
  MIN_SIZE_TEST: 448
  FORMAT: "RGB"
  CROP:
    ENABLED: True
    TYPE: "absolute_range"
    SIZE: (384, 600)
  LSJ_AUG:
    ENABLED: False
    IMAGE_SIZE: 768
    MIN_SCALE: 0.1
    MAX_SCALE: 2.0
DATALOADER:
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 8
OUTPUT_DIR: "./outputs/avism_SwinL_IN"