virtex/configs/detectron2/_base_faster_rcnn_R_50_C4_BN.yaml

# ----------------------------------------------------------------------------
# Train a Faster R-CNN with ResNet-50 and C4 backbone. This config follows
# Detectron2 format; and is unrelated with our VirTex configs. Params here
# replicate evaluation protocol as per MoCo (https://arxiv.org/abs/1911.05722).
# ----------------------------------------------------------------------------

INPUT:
  # Input format will always be RGB, consistent with torchvision.
  FORMAT: "RGB"
  MIN_SIZE_TRAIN: (640, 672, 704, 736, 768, 800)
  MIN_SIZE_TEST: 800

MODEL:
  META_ARCHITECTURE: "GeneralizedRCNN"

  # Train all layers end-to-end by default.
  BACKBONE:
    NAME: build_resnet_backbone
    FREEZE_AT: 0

  # Fine-tune with SyncBN.
  # STRIDE_IN_1X1 is False for torchvision-like models.
  RESNETS:
    DEPTH: 50
    NORM: SyncBN
    STRIDE_IN_1X1: False

  RPN:
    PRE_NMS_TOPK_TEST: 6000
    POST_NMS_TOPK_TEST: 1000

  # ROI head with extra BN layer after res5 stage.
  ROI_HEADS:
    NAME: "Res5ROIHeadsExtraNorm"

  # ImageNet color mean for torchvision-like models (RGB order).
  PIXEL_MEAN: [123.675, 116.280, 103.530]
  PIXEL_STD: [58.395, 57.120, 57.375]

SOLVER:
  # This is for 8 GPUs, apply linear scaling for 4 GPUs.
  IMS_PER_BATCH: 16
  BASE_LR: 0.02

TEST:
  PRECISE_BN:
    ENABLED: True

VERSION: 2