_BASE_: Base-Cityscapes-UnifiedSegmentation.yaml
MODEL:
  META_ARCHITECTURE: "OneFormer"
  SEM_SEG_HEAD:
    NAME: "OneFormerHead"
    IGNORE_VALUE: 255
    NUM_CLASSES: 19
    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
  ONE_FORMER:
    TRANSFORMER_DECODER_NAME: "ContrastiveMultiScaleMaskedTransformerDecoder"
    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
    CONTRASTIVE_WEIGHT: 0.5
    CONTRASTIVE_TEMPERATURE: 0.07
    HIDDEN_DIM: 256
    NUM_OBJECT_QUERIES: 150
    USE_TASK_NORM: True
    NHEADS: 8
    DROPOUT: 0.1
    DIM_FEEDFORWARD: 2048
    ENC_LAYERS: 0
    PRE_NORM: False
    ENFORCE_INPUT_PROJ: False
    SIZE_DIVISIBILITY: 32
    ENC_LAYERS: 0
    CLASS_DEC_LAYERS: 2
    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
  TEXT_ENCODER:
    WIDTH: 256
    CONTEXT_LENGTH: 77
    NUM_LAYERS: 6
    VOCAB_SIZE: 49408
    PROJ_NUM_LAYERS: 2
    N_CTX: 16
  TEST:
    SEMANTIC_ON: True
    INSTANCE_ON: True
    PANOPTIC_ON: True
    OVERLAP_THRESHOLD: 0.8
    OBJECT_MASK_THRESHOLD: 0.8
    TASK: "panoptic"
TEST:
  DETECTIONS_PER_IMAGE: 150