HRNet / models /Dual-HRNet /dual-hrnet.yaml
niobures's picture
HRNet (code, colab, models, paper)
f6c4e37 verified
OUTPUT_DIR: ''
LOG_DIR: ''
GPUS: [0,]
WORKERS: 4
PRINT_FREQ: 20
AUTO_RESUME: False
PIN_MEMORY: True
RANK: 0
# Cudnn related params
CUDNN:
BENCHMARK: True
DETERMINISTIC: False
ENABLED: True
# common params for NETWORK
MODEL:
NAME: 'dual-hrnet'
PRETRAINED: './Checkpoints/HRNet/hrnetv2_w32_imagenet_pretrained.pth'
USE_FPN: False
IS_DISASTER_PRED: False
IS_SPLIT_LOSS: True
FUSE_CONV_K_SIZE: 1
# high_resoluton_net related params for segmentation
EXTRA:
PRETRAINED_LAYERS: ['*']
STEM_INPLANES: 64
FINAL_CONV_KERNEL: 1
WITH_HEAD: True
STAGE1:
NUM_MODULES: 1
NUM_BRANCHES: 1
NUM_BLOCKS: [4]
NUM_CHANNELS: [64]
BLOCK: 'BOTTLENECK'
FUSE_METHOD: 'SUM'
STAGE2:
NUM_MODULES: 1
NUM_BRANCHES: 2
NUM_BLOCKS: [4, 4]
NUM_CHANNELS: [32, 64]
BLOCK: 'BASIC'
FUSE_METHOD: 'SUM'
STAGE3:
NUM_MODULES: 4
NUM_BRANCHES: 3
NUM_BLOCKS: [4, 4, 4]
NUM_CHANNELS: [32, 64, 128]
BLOCK: 'BASIC'
FUSE_METHOD: 'SUM'
STAGE4:
NUM_MODULES: 3
NUM_BRANCHES: 4
NUM_BLOCKS: [4, 4, 4, 4]
NUM_CHANNELS: [32, 64, 128, 256]
BLOCK: 'BASIC'
FUSE_METHOD: 'SUM'
#_C.MODEL.EXTRA= CN(new_allowed=True)
LOSS:
CLASS_BALANCE: True
# DATASET related params
DATASET:
NUM_CLASSES: 4
# training
TRAIN:
# Augmentation
FLIP: True
MULTI_SCALE: [0.8, 1.2]
CROP_SIZE: [512, 512]
LR_FACTOR: 0.1
LR_STEP: [90, 110]
LR: 0.05
EXTRA_LR: 0.001
OPTIMIZER: 'sgd'
MOMENTUM: 0.9
WD: 0.0001
NESTEROV: False
IGNORE_LABEL: -1
NUM_EPOCHS: 500
RESUME: False
BATCH_SIZE_PER_GPU: 16
SHUFFLE: True
# only using some training samples
NUM_SAMPLES: 0
CLASS_WEIGHTS: [0.4, 1.2, 1.2, 1.2]
# testing
TEST:
BATCH_SIZE_PER_GPU: 32
# only testing some samples
NUM_SAMPLES: 0
MODEL_FILE: ''
FLIP_TEST: False
MULTI_SCALE: False
CENTER_CROP_TEST: False
SCALE_LIST: [1]
# debug
DEBUG:
DEBUG: False
SAVE_BATCH_IMAGES_GT: False
SAVE_BATCH_IMAGES_PRED: False
SAVE_HEATMAPS_GT: False
SAVE_HEATMAPS_PRED: False