[자율주행 AI 챌린지] BaseLine 개선방안 결과/졸업작품 ppt영상

[김위제]

https://cafe.daum.net/SmartRobot/RoVa/2408

[자율주행 AI 챌린지] BaseLine 개선방안 결과/ 다음 개선방안

MODEL : centerpoint_pillars
DATASET : 2025 자율주행 AI 챌린지 dataset (128ch+64ch)
TEST CODE : 대회의 baseline code에서 val.py(2025)를 사용해서 확인
TEST DATASET : 2024 dataset에서 train dataset의 64ch data만 뽑아서 사용

---

centerpoint.yaml

---

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

DATA_CONFIG:

    _BASE_CONFIG_: cfgs/dataset_configs/custom_pillars_dataset.yaml

    POINT_CLOUD_RANGE: [-75.2, -75.2, -2.0, 75.2, 75.2, 4.0]

    DATA_PROCESSOR:

        -   NAME: mask_points_and_boxes_outside_range

            REMOVE_OUTSIDE_BOXES: True

        -   NAME: shuffle_points

            SHUFFLE_ENABLED: {

                'train': True,

                'test': True

            }

        -   NAME: transform_points_to_voxels

            VOXEL_SIZE: [ 0.2, 0.2, 6.0 ]

            MAX_POINTS_PER_VOXEL: 20

            MAX_NUMBER_OF_VOXELS: {

                'train': 250000, #150000

                'test': 250000 # 150000

            }

MODEL:

    NAME: CenterPoint

    VFE:

        NAME: PillarVFE

        WITH_DISTANCE: False

        USE_ABSLOTE_XYZ: True

        USE_NORM: True

        NUM_FILTERS: [ 64, 64 ]

    MAP_TO_BEV:

        NAME: PointPillarScatter

        NUM_BEV_FEATURES: 64

    BACKBONE_2D:

        NAME: BaseBEVBackbone

        LAYER_NUMS: [ 3, 5, 5 ]

        LAYER_STRIDES: [ 1, 2, 2 ]

        NUM_FILTERS: [ 64, 128, 256 ]

        UPSAMPLE_STRIDES: [ 1, 2, 4 ]

        NUM_UPSAMPLE_FILTERS: [ 128, 128, 128 ]

    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

        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: 1

            NUM_MAX_OBJS: 500

            GAUSSIAN_OVERLAP: 0.1

            MIN_RADIUS: 2

        LOSS_CONFIG:

            LOSS_WEIGHTS: {

                'cls_weight': 1.5, #클래스 불균형

                'loc_weight': 2.0,

                'code_weights': [1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0]

            }

            CLASS_WEIGHT : [1.5,2.5,2.0]

        POST_PROCESSING:

            SCORE_THRESH: 0.1

            POST_CENTER_LIMIT_RANGE: [-75.2, -75.2, -2.0, 75.2, 75.2, 4.0]

            MAX_OBJ_PER_SAMPLE: 500

            NMS_CONFIG:

                NMS_TYPE: nms_gpu

                NMS_THRESH: 0.7

                NMS_PRE_MAXSIZE: 4096

                NMS_POST_MAXSIZE: 500

    POST_PROCESSING:

        RECALL_THRESH_LIST: [0.3, 0.5, 0.7]

        EVAL‎_METRIC: waymo

OPTIMIZATION:

    BATCH_SIZE_PER_GPU: 1

    NUM_EPOCHS: 80

    OPTIMIZER: adam_onecycle

    LR: 0.003

    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

---

dataset.yaml

---

DATASET: 'CustomAvDataset'

DATA_PATH: '/mnt/c/data/custom_av'

POINT_CLOUD_RANGE: [-70.0, -70.0, -4.0, 70.0, 70.0, 4.0]

MAP_CLASS_TO_KITTI: {

    'Vehicle': 'Car',

    'Pedestrian': 'Pedestrian',

    'Cyclist': 'Cyclist',

}

DATA_SPLIT: {

    'train': train,

    'test': val

}

INFO_PATH: {

    'train': [custom_av_infos_train.pkl],

    'test': [custom_av_infos_val.pkl],

}

POINT_FEATURE_ENCODING: {

    encoding_type: absolute_coordinates_encoding,

    used_feature_list: ['x', 'y', 'z'],

    src_feature_list: ['x', 'y', 'z', 'intensity'],

}

DATA_AUGMENTOR:

    DISABLE_AUG_LIST: ['placeholder']

    AUG_CONFIG_LIST:

        - NAME: gt_sampling

          USE_ROAD_PLANE: False

          DB_INFO_PATH:

              - custom_av_dbinfos_train.pkl

          PREPARE: {

             filter_by_min_points: ['Vehicle:5', 'Pedestrian:5', 'Cyclist:5'],

          }

          SAMPLE_GROUPS: ['Vehicle:15', 'Pedestrian:10', 'Cyclist:10']

          NUM_POINT_FEATURES: 4

          DATABASE_WITH_FAKELIDAR: False

          REMOVE_EXTRA_WIDTH: [0.0, 0.0, 0.0]

          LIMIT_WHOLE_SCENE: True

        - NAME: random_world_flip

          ALONG_AXIS_LIST: ['x', 'y']

        - NAME: random_world_rotation

          WORLD_ROT_ANGLE: [-0.78539816, 0.78539816]

        - NAME: random_world_scaling

          WORLD_SCALE_RANGE: [0.95, 1.05]

DATA_PROCESSOR:

    - NAME: mask_points_and_boxes_outside_range

      REMOVE_OUTSIDE_BOXES: True

    - NAME: shuffle_points

      SHUFFLE_ENABLED: {

        'train': True,

        'test': True

      }

    - NAME: transform_points_to_voxels

      VOXEL_SIZE: [0.1, 0.1, 0.15]

      MAX_POINTS_PER_VOXEL: 5

      MAX_NUMBER_OF_VOXELS: {

        'train': 150000,

        'test': 150000

      }

---

리더보드

---

리더보드 TEST	baseline	개선방안
VEHICLE_AP/L1	0.5973	0.4786
VEHICLE_AP/L2	0.5687	0.4550
PEDESTRIAN_AP/L1	0.3480	0.1900
PEDESTRIAN_AP/L2	0.3372	0.1839
CYCLIST_AP/L1	0.3448	0.3284
CYCLIST_AP/L2	0.3351	0.3191

---

결론

---

baseline에서  BATCH_SIZE_PER_GPU: 1인 상태에서 80epoch으로는 제공해준 프리트레인 모델과 같은 성능을 낼수 없을 것 같다.
또한 voxelsize를0.25에서 0.2로 줄였는데 centerpoint_pillar에서는 voxelsize를 줄이는방법이 추론시간도 올라간것을 보아
voxelsize를 1등이 했던 방법이 아닌것 같기도 하고 성능을 방법이 될수 없는것 같다.

졸업작품 영상

---

원시 포인트 클라우드

Voxel_RCNN

4분할 영상

이러한 형식으로 1분정도 영상을 만들었습니다.

[Sungryul Lee]

영상추가, 버드아이뷰 보다는 자동차의 진행방향 시점으로 변경하여 3차원 박스가 보이도록 할것
참고
https://www.youtube.com/watch?v=eDbWZ1Q7fNc

[김위제]

넵 알겠습니다.