LoGoNet-1f.yaml

CLASS_NAMES: ['Vehicle', 'Pedestrian', 'Cyclist']

DATA_CONFIG:
    _BASE_CONFIG_: cfgs/det_dataset_cfgs/waymo_one_sweep_mm.yaml
    
MODEL:
    NAME: CenterPointMM
    SECOND_STAGE: True
    # remember to use "--pretrained_model" to load 1-stage checkpoint
    # and use "--fix_pretrained_weights" to fix params

    VFE:
        NAME: DynamicMeanVFE
        
    IMAGE_BACKBONE:
        NAME: MMDETFPN
        FREEZE_IMGBACKBONE: True
        # IMGPRETRAINED_MODEL: '../output/det_model_cfgs/pretrained_ckpt/swin_tiny_patch4_window7_224.pth'
        IFN:
            ARGS: {
                "feat_extract_layer": ["layer1"],
            }
            CHANNEL_REDUCE: {
                "in_channels": [256],
                "out_channels": [64],
                "kernel_size": [1],
                "stride": [1],
                "bias": [False]
            }
        IMGCONFIG: {
            type: 'SwinTransformer',
            embed_dim: 96,
            depths: [2, 2, 6, 2],
            num_heads: [3, 6, 12, 24],
            window_size: 7,
            mlp_ratio: 4,
            qkv_bias: True,
            qk_scale: None,
            drop_rate: 0.,
            attn_drop_rate: 0.,
            drop_path_rate: 0.2,
            ape: False,
            patch_norm: True,
            out_indices: [0, 1, 2, 3],
            use_checkpoint: False,
        }
        FPNCONFIG: {
            type: 'FPN',
            in_channels: [96, 192, 384, 768],
            out_channels: 256,
            num_outs: 5
        }

    BACKBONE_3D:
        NAME: Backbone3D_align

    MAP_TO_BEV:
        NAME: HeightCompression
        NUM_BEV_FEATURES: 256

    BACKBONE_2D:
        NAME: Backbone2D
        LAYER_NUMS: [5, 5]
        LAYER_STRIDES: [1, 2]
        NUM_FILTERS: [128, 256]
        UPSAMPLE_STRIDES: [1, 2]
        NUM_UPSAMPLE_FILTERS: [256, 256]

    DENSE_HEAD:
        NAME: CenterHead
        CLASS_AGNOSTIC: False

        CLASS_NAMES_EACH_HEAD: [
            [ 'Vehicle', 'Pedestrian', 'Cyclist' ]
        ]
        
        SHARED_CONV_CHANNEL: 64
        USE_BIAS_BEFORE_NORM: True
        NUM_HM_CONV: 2
        IOU_WEIGHT: 1
        SEPARATE_HEAD_CFG:
            HEAD_ORDER: [ 'center', 'center_z', 'dim', 'rot']
            HEAD_DICT: {
                'center': { 'out_channels': 2, 'num_conv': 2 },
                'center_z': { 'out_channels': 1, 'num_conv': 2 },
                'dim': { 'out_channels': 3, 'num_conv': 2 },
                'rot': { 'out_channels': 2, 'num_conv': 2 }
            }

        TARGET_ASSIGNER_CONFIG:
            FEATURE_MAP_STRIDE: 8
            NUM_MAX_OBJS: 500
            GAUSSIAN_OVERLAP: 0.1
            MIN_RADIUS: 2

        LOSS_CONFIG:
            LOSS_WEIGHTS: {
                'cls_weight': 1.0,
                'loc_weight': 2.0,
                'code_weights': [ 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0 ]
            }

        POST_PROCESSING:
            SCORE_THRESH: 0.1
            POST_CENTER_LIMIT_RANGE: [-80, -80, -10.0, 80, 80, 10.0]
            MAX_OBJ_PER_SAMPLE: 500
            NMS_CONFIG:
                NMS_TYPE: nms_gpu
                NMS_THRESH: 0.7
                NMS_PRE_MAXSIZE: 4096
                NMS_POST_MAXSIZE: 500
    ROI_HEAD:
        NAME: LoGoHead
        CLASS_AGNOSTIC: True

        SHARED_FC: [256, 256]
        CLS_FC: [256, 256]
        REG_FC: [256, 256]
        DP_RATIO: 0.3

        NMS_CONFIG:
            TRAIN:
                NMS_TYPE: nms_gpu
                MULTI_CLASSES_NMS: False
                NMS_PRE_MAXSIZE: 9000
                NMS_POST_MAXSIZE: 512
                NMS_THRESH: 0.8
            TEST:
                NMS_TYPE: nms_gpu
                MULTI_CLASSES_NMS: False
                NMS_PRE_MAXSIZE: 1024
                NMS_POST_MAXSIZE: 512
                NMS_THRESH: 0.7

        VOXEL_AGGREGATION:
            NUM_FEATURES: [64, 128]
            FEATURE_LOCATIONS: [x_conv3, x_conv4]
        ROI_GRID_POOL:
            FEATURE_LOCATIONS: [x_conv3, x_conv4]
            GRID_SIZE: 6
            POOL_LAYERS:
                x_conv3:
                    MLPS: [[32, 32], [32, 32]]
                    POOL_RADIUS: [0.8, 1.2]
                    NSAMPLE: [16, 16]
                    POOL_METHOD: max_pool
                    USE_DENSITY: True
                x_conv4:
                    MLPS: [[64, 64], [64, 64]]
                    POOL_RADIUS: [1.2, 2.4]
                    NSAMPLE: [16, 16]
                    POOL_METHOD: max_pool
                    USE_DENSITY: True
            FUSION:
                NAME: VoxelWithPointProjectionV2
                FUSE_MODE: 'crossattention_deform'
                CAMERAS: ['camera_0', 'camera_1', 'camera_2', 'camera_3', 'camera_4']
                IMAGE_SCALE: 0.5
                DEPTH_THRES: {'camera_0': 0, 'camera_1': 0, 'camera_2': 0, 'camera_3': 0, 'camera_4': 0}
                LAYER_CHANNEL: {'layer1': 64}
                MID_CHANNELS: 64
                FUSE_OUT: True
            ATTENTION:
                ENABLED: True
                NUM_FEATURES: 192
                NUM_HEADS: 1
                NUM_HIDDEN_FEATURES: 128
                NUM_LAYERS: 1
                POSITIONAL_ENCODER: density_grid_points
                MAX_NUM_BOXES: 20
                DROPOUT: 0.1
                COMBINE: True
                MASK_EMPTY_POINTS: True
            DENSITYQUERY:
                ENABLED: True
                NUM_FEATURES: 192
                NUM_HIDDEN_FEATURES: 128
                POSITIONAL_ENCODER: density_grid_points
                MAX_NUM_BOXES: 20
                DROPOUT: 0.1
                COMBINE: True
                MASK_EMPTY_POINTS: True
        TARGET_CONFIG:
            BOX_CODER: ResidualCoder
            ROI_PER_IMAGE: 128
            FG_RATIO: 0.5

            SAMPLE_ROI_BY_EACH_CLASS: True
            CLS_SCORE_TYPE: roi_iou

            CLS_FG_THRESH: 0.75
            CLS_BG_THRESH: 0.25
            CLS_BG_THRESH_LO: 0.1
            HARD_BG_RATIO: 0.8

            REG_FG_THRESH: 0.55

        LOSS_CONFIG:
            CLS_LOSS: BinaryCrossEntropy
            REG_LOSS: smooth-l1
            CORNER_LOSS_REGULARIZATION: True
            LOSS_WEIGHTS: {
                'rcnn_cls_weight': 1.0,
                'rcnn_reg_weight': 1.0,
                'rcnn_corner_weight': 1.0,
                'code_weights': [1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0]
            }

    POST_PROCESSING:
        RECALL_THRESH_LIST: [0.3, 0.5, 0.7]
        SCORE_THRESH: 0.1
        OUTPUT_RAW_SCORE: False

        EVAL_METRIC: waymo

        NMS_CONFIG:
            MULTI_CLASSES_NMS: False
            NMS_TYPE: nms_gpu
            NMS_THRESH: 0.7
            NMS_PRE_MAXSIZE: 4096
            NMS_POST_MAXSIZE: 500

OPTIMIZATION:
    BATCH_SIZE_PER_GPU: 4
    NUM_EPOCHS: 6

    OPTIMIZER: adam_onecycle
    LR: 0.001
    WEIGHT_DECAY: 0.01
    MOMENTUM: 0.9

    MOMS: [0.95, 0.85]
    PCT_START: 0.4
    DIV_FACTOR: 10
    DECAY_STEP_LIST: [35, 45]
    LR_DECAY: 0.1
    LR_CLIP: 0.0000001

    LR_WARMUP: False
    WARMUP_EPOCH: 1

    GRAD_NORM_CLIP: 10