resnet101_8x8f_baseline.yaml

MODEL:
  WEIGHT: "data/models/pretrained_models/SlowFast-ResNet101-8x8.pth"
  BACKBONE:
    CONV_BODY: "Slowfast-Resnet101"
    FROZEN_BN: True
    SLOWFAST:
      BETA: 0.125
      LATERAL: "tconv"
      SLOW:
        ACTIVE: True
        CONV3_NONLOCAL: False
        CONV4_NONLOCAL: False
      FAST:
        ACTIVE: True
        CONV3_NONLOCAL: False
        CONV4_NONLOCAL: False
  NONLOCAL:
    USE_ZERO_INIT_CONV: False
    BN_INIT_GAMMA: 0.0
    FROZEN_BN: True
  ROI_ACTION_HEAD:
    FEATURE_EXTRACTOR: "2MLPFeatureExtractor"
    POOLER_TYPE: "align3d"
    MEAN_BEFORE_POOLER: True
    POOLER_RESOLUTION: 7
    POOLER_SCALE: 0.0625
    POOLER_SAMPLING_RATIO: 0
    NUM_CLASSES: 80
    PROPOSAL_PER_CLIP: 10
    DROPOUT_RATE: 0.2
  IA_STRUCTURE:
    ACTIVE: False
INPUT:
  FRAME_NUM: 64
  FRAME_SAMPLE_RATE: 1
  TAU: 8
  ALPHA: 4
  SLOW_JITTER: True
  COLOR_JITTER: True
DATASETS:
  TRAIN: ("ava_video_train_v2.2",)
  TEST: ("ava_video_val_v2.2",)
DATALOADER:
  NUM_WORKERS: 4
  SIZE_DIVISIBILITY: 16
SOLVER:
  BASE_LR: 0.0004
  WARMUP_FACTOR: 0.25
  BIAS_LR_FACTOR: 2
  WEIGHT_DECAY: 1e-7
  STEPS: (50000, 70000)
  WARMUP_ITERS: 2000
  MAX_ITER: 90000
  CHECKPOINT_PERIOD: 10000
  EVAL_PERIOD: 10000
  VIDEOS_PER_BATCH: 16
TEST:
  BOX_THRESH: 0.8
  ACTION_THRESH: 0.
  VIDEOS_PER_BATCH: 16
OUTPUT_DIR: "data/output/resnet101_8x8f_baseline"