detection / VLP /FasterRCNN_X_101_64x4d_FPN_2x_config.yaml
fuliucansheng's picture
add vlp detection yaml & weight
419243b
MODEL:
META_ARCHITECTURE: "GeneralizedRCNN"
MASK_ON: False
BACKBONE:
NAME: "build_resnet_fpn_backbone"
RESNETS:
OUT_FEATURES: ["res2", "res3", "res4", "res5"]
STRIDE_IN_1X1: False # this is a C2 model
NUM_GROUPS: 64
WIDTH_PER_GROUP: 4
DEPTH: 101
FPN:
IN_FEATURES: ["res2", "res3", "res4", "res5"]
OUT_CHANNELS: 512
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
NMS_THRESH: 0.7
ROI_HEADS:
NAME: "StandardROIHeads"
IN_FEATURES: ["p2", "p3", "p4", "p5"]
NUM_CLASSES: 1601
NMS_THRESH_TEST: 0.3
ROI_BOX_HEAD:
NAME: "FastRCNNConvFCHead"
NUM_CONV: 0
NUM_FC: 2
FC_DIM: 2048
POOLER_RESOLUTION: 7
POOLER_SAMPLING_RATIO: 2
POOLER_TYPE: "ROIAlign"
VERSION: 2