multi-class version

[kyle263]

Hi, I followed the difference between voxel_rcnn_car.yaml and voxel_rcnn_3classes.yaml to change the VirConv-L.yaml as below, but after training still get 0 in pedestrian and cyclist, could you please release the multi-class version?

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

DATA_CONFIG:
    _BASE_CONFIG_: cfgs/dataset_configs/kitti_dataset.yaml
    DATASET: 'KittiDatasetMM'
    INPUT_DISCARD_RATE: 0.8   # actually discard more than 90% virtual points of RGB image,
                              # as we only saved less than 50% RGB points during depth2points conversion.
    LATER_FUSION: False
    MM_PATH: 'velodyne_depth'
    USE_VAN: True

    DATA_SPLIT: {
        'train': train,
        'test': val
    }

    INFO_PATH: {
        'train': [kitti_infos_train.pkl],
        'test': [kitti_infos_val.pkl],
    }

    DATA_AUGMENTOR:
        DISABLE_AUG_LIST: ['placeholder']
        AUG_CONFIG_LIST:
            - NAME: gt_sampling
              USE_ROAD_PLANE: True
              DB_INFO_PATH:
                  - kitti_dbinfos_train_mm.pkl
              PREPARE: {
                  filter_by_min_points: ['Car:5', 'Pedestrian:5', 'Cyclist:5'],
                  filter_by_difficulty: [-1],
              }

              SAMPLE_GROUPS: ['Car:10', 'Pedestrian:10', 'Cyclist:10']
              NUM_POINT_FEATURES: 8
              DATABASE_WITH_FAKELIDAR: False
              REMOVE_EXTRA_WIDTH: [0.0, 0.0, -0.2]
              LIMIT_WHOLE_SCENE: False

            - NAME: da_sampling
              USE_ROAD_PLANE: True
              DB_INFO_PATH:
                - kitti_dbinfos_train_mm.pkl
              PREPARE: {
                filter_by_min_points: ['Car:5', 'Pedestrian:5', 'Cyclist:5'],
                filter_by_difficulty: [-1],
              }

              SAMPLE_GROUPS: ['Car:10', 'Pedestrian:10', 'Cyclist:10']

              MIN_SAMPLING_DIS: 0
              MAX_SAMPLING_DIS: 20
              OCCLUSION_NOISE: 0.2
              OCCLUSION_OFFSET: 2.
              SAMPLING_METHOD: 'LiDAR-aware'
              VERT_RES: 0.006
              HOR_RES: 0.003

              NUM_POINT_FEATURES: 8
              DATABASE_WITH_FAKELIDAR: False
              REMOVE_EXTRA_WIDTH: [0.0, 0.0, -0.2]
              LIMIT_WHOLE_SCENE: False

            - NAME: random_local_noise
              LOCAL_ROT_RANGE: [-0.78539816, 0.78539816]
              TRANSLATION_STD: [1.0, 1.0, 0.5]
              GLOBAL_ROT_RANGE: [0.0, 0.0]
              EXTRA_WIDTH: [0.2, 0.2, 0.]

            - NAME: random_world_rotation
              WORLD_ROT_ANGLE: [-0.78539816, 0.78539816]

            - NAME: random_world_flip
              ALONG_AXIS_LIST: ['x']

            - NAME: random_world_scaling
              WORLD_SCALE_RANGE: [0.95, 1.05]

            - NAME: random_local_pyramid_aug
              DROP_PROB: 0.25
              SPARSIFY_PROB: 0.05
              SPARSIFY_MAX_NUM: 50
              SWAP_PROB: 0.1
              SWAP_MAX_NUM: 50

    POINT_FEATURE_ENCODING: {
        encoding_type: absolute_coordinates_encoding_mm,
        used_feature_list: ['x', 'y', 'z', 'intensity'],
        src_feature_list: ['x', 'y', 'z', 'intensity'],
        num_features: 8
    }

    DATA_PROCESSOR:
        - NAME: mask_points_and_boxes_outside_range
          REMOVE_OUTSIDE_BOXES: True

        - NAME: shuffle_points
          SHUFFLE_ENABLED: {
            'train': True,
            'test': False
          }

        - NAME: transform_points_to_voxels
          VOXEL_SIZE: [0.05, 0.05, 0.05]
          LIDAR_FIRST: True
          MAX_POINTS_PER_VOXEL: 5
          MAX_NUMBER_OF_VOXELS: {
            'train': 40000, # using 40000 voxels is enough after input discard
            'test': 40000   # The 40000 voxels include both LiDAR and virtual points
          }

MODEL:
    NAME: VoxelRCNN

    VFE:
        NAME: MeanVFE
        MODEL: 'max'

    BACKBONE_3D:
        NAME: VirConvL8x
        NUM_FILTERS: [16, 32, 64, 64]
        RETURN_NUM_FEATURES_AS_DICT: True
        OUT_FEATURES: 64
        LAYER_DISCARD_RATE: 0.1

    MAP_TO_BEV:
        NAME: HeightCompression
        NUM_BEV_FEATURES: 256

    BACKBONE_2D:
        NAME: BaseBEVBackbone

        LAYER_NUMS: [4, 4]
        LAYER_STRIDES: [1, 2]
        NUM_FILTERS: [64, 128]
        UPSAMPLE_STRIDES: [1, 2]
        NUM_UPSAMPLE_FILTERS: [128, 128]

    DENSE_HEAD:
        NAME: AnchorHeadSingle
        CLASS_AGNOSTIC: False

        USE_DIRECTION_CLASSIFIER: True
        DIR_OFFSET: 0.78539
        DIR_LIMIT_OFFSET: 0.0
        NUM_DIR_BINS: 2

        ANCHOR_GENERATOR_CONFIG: [
            {
                'class_name': 'Car',
                'anchor_sizes': [[3.9, 1.6, 1.56]],
                'anchor_rotations': [0, 1.57],
                'anchor_bottom_heights': [-1.78],
                'align_center': False,
                'feature_map_stride': 8,
                'matched_threshold': 0.6,
                'unmatched_threshold': 0.45
            },
            {
                'class_name': 'Pedestrian',
                'anchor_sizes': [[0.8, 0.6, 1.73]],
                'anchor_rotations': [0, 1.57],
                'anchor_bottom_heights': [-0.6],
                'align_center': False,
                'feature_map_stride': 8,
                'matched_threshold': 0.5,
                'unmatched_threshold': 0.35
            },
            {
                'class_name': 'Cyclist',
                'anchor_sizes': [[1.76, 0.6, 1.73]],
                'anchor_rotations': [0, 1.57],
                'anchor_bottom_heights': [-0.6],
                'align_center': False,
                'feature_map_stride': 8,
                'matched_threshold': 0.5,
                'unmatched_threshold': 0.35
            }
        ]
        TARGET_ASSIGNER_CONFIG:
            NAME: AxisAlignedTargetAssigner
            POS_FRACTION: -1.0
            SAMPLE_SIZE: 512
            NORM_BY_NUM_EXAMPLES: False
            MATCH_HEIGHT: False
            BOX_CODER: ResidualCoder

        LOSS_CONFIG:
            LOSS_WEIGHTS: {
                'cls_weight': 1.0,
                'loc_weight': 2.0,
                'dir_weight': 0.2,
                'code_weights': [1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0]
            }

    ROI_HEAD:
        NAME: TEDMHead
        CLASS_AGNOSTIC: True
        ROT_NUM: 1 # set 1 to simply keep a similar architecture as Voxel-RCNN

        SHARED_FC: [256, 256]
        CLS_FC: [256, 256]
        REG_FC: [256, 256]
        DP_RATIO: 0.01

        NMS_CONFIG:
            TRAIN:
                NMS_TYPE: nms_gpu
                MULTI_CLASSES_NMS: False
                NMS_PRE_MAXSIZE: 4000
                NMS_POST_MAXSIZE: 512
                NMS_THRESH: 0.8
            TEST:
                NMS_TYPE: nms_gpu
                MULTI_CLASSES_NMS: False
                USE_FAST_NMS: True
                SCORE_THRESH: 0.0
                NMS_PRE_MAXSIZE: 2000
                NMS_POST_MAXSIZE: 50
                NMS_THRESH: 0.75

        ROI_GRID_POOL:
            FEATURES_SOURCE: ['x_conv3','x_conv4']
            PRE_MLP: True
            GRID_SIZE: 6
            POOL_LAYERS:
                x_conv3:
                    MLPS: [[32, 32], [32, 32]]
                    QUERY_RANGES: [[2, 2, 2], [4, 4, 4]]
                    POOL_RADIUS: [0.4, 0.8]
                    NSAMPLE: [16, 16]
                    POOL_METHOD: max_pool
                x_conv4:
                    MLPS: [[32, 32], [32, 32]]
                    QUERY_RANGES: [[2, 2, 2], [4, 4, 4]]
                    POOL_RADIUS: [0.8, 1.6]
                    NSAMPLE: [16, 16]
                    POOL_METHOD: max_pool

        ROI_GRID_POOL_MM:
            FEATURES_SOURCE: ['x_conv3','x_conv4']
            PRE_MLP: True
            GRID_SIZE: 4
            POOL_LAYERS:
                x_conv3:
                    MLPS: [[32, 32], [32, 32]]
                    QUERY_RANGES: [[2, 2, 2], [4, 4, 4]]
                    POOL_RADIUS: [0.4, 0.8]
                    NSAMPLE: [16, 16]
                    POOL_METHOD: max_pool
                x_conv4:
                    MLPS: [[32, 32], [32, 32]]
                    QUERY_RANGES: [[2, 2, 2], [4, 4, 4]]
                    POOL_RADIUS: [0.8, 1.6]
                    NSAMPLE: [16, 16]
                    POOL_METHOD: max_pool

        TARGET_CONFIG:
            BOX_CODER: ResidualCoder
            STAGE0:
                ROI_PER_IMAGE: 160
                FG_RATIO: 0.5
                SAMPLE_ROI_BY_EACH_CLASS: True
                CLS_SCORE_TYPE: roi_iou_x
                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]
                ENABLE_HARD_SAMPLING: True
                HARD_SAMPLING_THRESH: [0.5]
                HARD_SAMPLING_RATIO: [0.5]

        LOSS_CONFIG:
            CLS_LOSS: BinaryCrossEntropy
            REG_LOSS: smooth-l1
            CORNER_LOSS_REGULARIZATION: True
            GRID_3D_IOU_LOSS: False
            LOSS_WEIGHTS: {
                'rcnn_cls_weight': 1.0,
                'rcnn_reg_weight': 1.0,
                'rcnn_corner_weight': 1.0,
                'rcnn_iou3d_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: 0.5
        OUTPUT_RAW_SCORE: False
        EVAL_METRIC: kitti
        NMS_CONFIG:
            MULTI_CLASSES_NMS: False
            NMS_TYPE: nms_gpu
            NMS_THRESH: 0.1
            NMS_PRE_MAXSIZE: 4096
            NMS_POST_MAXSIZE: 500

OPTIMIZATION:
    BATCH_SIZE_PER_GPU: 2
    NUM_EPOCHS: 50

    OPTIMIZER: adam_onecycle
    LR: 0.01
    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

[hailanyi]

please refer to issue 8