Upload folder using huggingface_hub

SiMO_AF/config.yaml

@@ -0,0 +1,339 @@
+add_data_extension:
+- bev_visibility.png
+cav_lidar_range: &id002
+- -51.2
+- -51.2
+- -3
+- 51.2
+- 51.2
+- 1
+comm_range: 70
+comment: ''
+data_augment:
+- ALONG_AXIS_LIST:
+  - x
+  NAME: random_world_flip
+- NAME: random_world_rotation
+  WORLD_ROT_ANGLE:
+  - -0.78539816
+  - 0.78539816
+- NAME: random_world_scaling
+  WORLD_SCALE_RANGE:
+  - 0.95
+  - 1.05
+fusion:
+  args:
+    data_aug_conf: &id004
+      H: 600
+      Ncams: 4
+      W: 800
+      bot_pct_lim:
+      - 0.0
+      - 0.05
+      cams:
+      - camera0
+      - camera1
+      - camera2
+      - camera3
+      final_dim:
+      - 384
+      - 512
+      rand_flip: false
+      resize_lim:
+      - 0.65
+      - 0.7
+      rot_lim:
+      - -3.6
+      - 3.6
+    grid_conf: &id003
+      ddiscr:
+      - 2
+      - 50
+      - 48
+      mode: LID
+      xbound:
+      - -51.2
+      - 51.2
+      - 0.4
+      ybound:
+      - -51.2
+      - 51.2
+      - 0.4
+      zbound:
+      - -10
+      - 10
+      - 20.0
+    proj_first: false
+  core_method: intermediate
+  dataset: opv2v
+heter:
+  modality_setting:
+    m1:
+      core_method: point_pillar
+      freeze: true
+      model_dir: opencood/logs/HeterBaseline_opv2v_lidar_attfuse_2024_12_02_22_07_35
+      sensor_type: lidar
+    m2:
+      core_method: lift_splat_shoot
+      freeze: true
+      model_dir: opencood/logs/official_checkpoints/HeterBaseline_opv2v_camera_attfuse_2023_08_08_16_50_01
+      sensor_type: camera
+input_source:
+- lidar
+- camera
+- depth
+label_type: lidar
+loss:
+  args:
+    cls:
+      alpha: 0.25
+      gamma: 2.0
+      type: SigmoidFocalLoss
+      weight: 1.0
+    depth:
+      weight: 1.0
+    dir:
+      args: &id001
+        anchor_yaw: &id006
+        - 0
+        - 90
+        dir_offset: 0.7853
+        num_bins: 2
+      type: WeightedSoftmaxClassificationLoss
+      weight: 0.2
+    pos_cls_weight: 2.0
+    reg:
+      codewise: true
+      sigma: 3.0
+      type: WeightedSmoothL1Loss
+      weight: 2.0
+  core_method: point_pillar_loss
+lr_scheduler:
+  core_method: multistep
+  gamma: 0.1
+  step_size:
+  - 10
+  - 40
+model:
+  args:
+    anchor_number: 2
+    att:
+      feat_dim: 256
+    dir_args: *id001
+    head_freeze: true
+    in_head: 256
+    lamma:
+      dim: 256
+      feat_dim: 256
+      feature_stride: 2
+      freeze: false
+      heads: 2
+      lidar_drop_ratio: 0.5
+      random_drop: false
+      single_mode: 'camera'
+    lidar_range: *id002
+    m1:
+      aligner_args:
+        args:
+          dim: 256
+          num_of_blocks: 3
+        core_method: convnext
+        freeze: true
+        spatial_align: false
+      backbone_args:
+        freeze: true
+        layer_nums:
+        - 3
+        - 5
+        - 8
+        layer_strides:
+        - 2
+        - 2
+        - 2
+        num_filters:
+        - 64
+        - 128
+        - 256
+        num_upsample_filter:
+        - 128
+        - 128
+        - 128
+        upsample_strides:
+        - 1
+        - 2
+        - 4
+      core_method: point_pillar
+      encoder_args:
+        freeze: true
+        lidar_range: *id002
+        pillar_vfe:
+          num_filters:
+          - 64
+          use_absolute_xyz: true
+          use_norm: true
+          with_distance: false
+        point_pillar_scatter:
+          grid_size: !!python/object/apply:numpy.core.multiarray._reconstruct
+            args:
+            - !!python/name:numpy.ndarray ''
+            - !!python/tuple
+              - 0
+            - !!binary |
+              Yg==
+            state: !!python/tuple
+            - 1
+            - !!python/tuple
+              - 3
+            - !!python/object/apply:numpy.dtype
+              args:
+              - i8
+              - false
+              - true
+              state: !!python/tuple
+              - 3
+              - <
+              - null
+              - null
+              - null
+              - -1
+              - -1
+              - 0
+            - false
+            - !!binary |
+              AAEAAAAAAAAAAQAAAAAAAAEAAAAAAAAA
+          num_features: 64
+        voxel_size: &id005
+        - 0.4
+        - 0.4
+        - 4
+      freeze: true
+      model_dir: opencood/logs/HeterBaseline_opv2v_lidar_attfuse_2024_12_02_22_07_35
+      sensor_type: lidar
+      shrink_header:
+        dim:
+        - 256
+        freeze: true
+        input_dim: 384
+        kernal_size:
+        - 3
+        padding:
+        - 1
+        stride:
+        - 1
+    m2:
+      aligner_args:
+        args:
+          dim: 256
+          num_of_blocks: 3
+        core_method: convnext
+        freeze: true
+        spatial_align: false
+      backbone_args:
+        freeze: true
+        inplanes: 128
+        layer_nums:
+        - 3
+        - 5
+        - 8
+        layer_strides:
+        - 2
+        - 2
+        - 2
+        num_filters:
+        - 64
+        - 128
+        - 256
+        num_upsample_filter:
+        - 128
+        - 128
+        - 128
+        upsample_strides:
+        - 1
+        - 2
+        - 4
+      camera_mask_args:
+        cav_lidar_range: *id002
+        grid_conf: *id003
+      core_method: lift_splat_shoot
+      encoder_args:
+        anchor_number: 2
+        camera_encoder: EfficientNet
+        data_aug_conf: *id004
+        depth_supervision: true
+        freeze: true
+        grid_conf: *id003
+        img_downsample: 8
+        img_features: 128
+        use_depth_gt: false
+      freeze: true
+      model_dir: opencood/logs/official_checkpoints/HeterBaseline_opv2v_camera_attfuse_2023_08_08_16_50_01
+      sensor_type: camera
+      shrink_header:
+        dim:
+        - 256
+        freeze: true
+        input_dim: 384
+        kernal_size:
+        - 3
+        padding:
+        - 1
+        stride:
+        - 1
+    ma_fusion_method: att
+    mm_fusion_method: lamma3
+    single_modality: 'camera'
+    voxel_size: *id005
+  core_method: point_pillar_lss_lamma2_attfuse
+name: opv2v_lidarcamera_lamma3_attfuse
+noise_setting: !!python/object/apply:collections.OrderedDict
+- - - add_noise
+    - false
+optimizer:
+  args:
+    eps: 1.0e-10
+    weight_decay: 0.0001
+  core_method: Adam
+  lr: 0.0002
+postprocess:
+  anchor_args:
+    D: 1
+    H: 256
+    W: 256
+    cav_lidar_range: *id002
+    feature_stride: 2
+    h: 1.56
+    l: 3.9
+    num: 2
+    r: *id006
+    vd: 4
+    vh: 0.4
+    vw: 0.4
+    w: 1.6
+  core_method: VoxelPostprocessor
+  dir_args: *id001
+  gt_range: *id002
+  max_num: 150
+  nms_thresh: 0.15
+  order: hwl
+  target_args:
+    neg_threshold: 0.45
+    pos_threshold: 0.6
+    score_threshold: 0.2
+preprocess:
+  args:
+    max_points_per_voxel: 32
+    max_voxel_test: 70000
+    max_voxel_train: 32000
+    voxel_size: *id005
+  cav_lidar_range: *id002
+  core_method: SpVoxelPreprocessor
+root_dir: data/OPV2V/train
+test_dir: data/OPV2V/test
+train_params:
+  batch_size: 1
+  epoches: 40
+  eval_freq: 2
+  max_cav: 5
+  save_freq: 2
+validate_dir: data/OPV2V/validate
+yaml_parser: load_general_params

SiMO_PF/config.yaml

@@ -0,0 +1,309 @@
+add_data_extension:
+- bev_visibility.png
+cav_lidar_range: &id002
+- -51.2
+- -51.2
+- -3
+- 51.2
+- 51.2
+- 1
+comm_range: 70
+comment: ''
+data_augment:
+- ALONG_AXIS_LIST:
+  - x
+  NAME: random_world_flip
+- NAME: random_world_rotation
+  WORLD_ROT_ANGLE:
+  - -0.78539816
+  - 0.78539816
+- NAME: random_world_scaling
+  WORLD_SCALE_RANGE:
+  - 0.95
+  - 1.05
+fusion:
+  args:
+    data_aug_conf: &id004
+      H: 600
+      Ncams: 4
+      W: 800
+      bot_pct_lim:
+      - 0.0
+      - 0.05
+      cams:
+      - camera0
+      - camera1
+      - camera2
+      - camera3
+      final_dim:
+      - 384
+      - 512
+      rand_flip: false
+      resize_lim:
+      - 0.65
+      - 0.7
+      rot_lim:
+      - -3.6
+      - 3.6
+    grid_conf: &id003
+      ddiscr:
+      - 2
+      - 50
+      - 48
+      mode: LID
+      xbound:
+      - -51.2
+      - 51.2
+      - 0.4
+      ybound:
+      - -51.2
+      - 51.2
+      - 0.4
+      zbound:
+      - -10
+      - 10
+      - 20.0
+    proj_first: false
+  core_method: intermediate
+  dataset: opv2v
+heter:
+  modality_setting:
+    m1:
+      core_method: point_pillar
+      sensor_type: lidar
+    m2:
+      core_method: lift_splat_shoot
+      sensor_type: camera
+input_source:
+- lidar
+- camera
+- depth
+label_type: lidar
+loss:
+  args:
+    cls:
+      alpha: 0.25
+      gamma: 2.0
+      type: SigmoidFocalLoss
+      weight: 1.0
+    depth:
+      weight: 1.0
+    dir:
+      args: &id001
+        anchor_yaw: &id006
+        - 0
+        - 90
+        dir_offset: 0.7853
+        num_bins: 2
+      type: WeightedSoftmaxClassificationLoss
+      weight: 0.2
+    pos_cls_weight: 2.0
+    reg:
+      codewise: true
+      sigma: 3.0
+      type: WeightedSmoothL1Loss
+      weight: 2.0
+  core_method: point_pillar_loss
+lr_scheduler:
+  core_method: multistep
+  gamma: 0.1
+  step_size:
+  - 15
+  - 40
+model:
+  args:
+    anchor_number: 2
+    dir_args: *id001
+    fusion_backbone:
+      anchor_number: 2
+      freeze: true
+      layer_nums:
+      - 3
+      - 5
+      - 8
+      layer_strides:
+      - 1
+      - 2
+      - 2
+      num_filters:
+      - 64
+      - 128
+      - 256
+      num_upsample_filter:
+      - 128
+      - 128
+      - 128
+      resnext: true
+      upsample_strides:
+      - 1
+      - 2
+      - 4
+    head_freeze: true
+    in_head: 256
+    lamma:
+      dim: 128
+      feat_dim: 64
+      feature_stride: 2
+      freeze: false
+      heads: 2
+      lidar_drop_ratio: 0.5
+      random_drop: false
+      single_mode: false
+    lidar_range: *id002
+    m1:
+      aligner_args:
+        core_method: identity
+        freeze: true
+      backbone_args:
+        freeze: true
+        layer_nums:
+        - 3
+        layer_strides:
+        - 2
+        num_filters:
+        - 64
+      core_method: point_pillar
+      encoder_args:
+        freeze: true
+        lidar_range: *id002
+        pillar_vfe:
+          num_filters:
+          - 64
+          use_absolute_xyz: true
+          use_norm: true
+          with_distance: false
+        point_pillar_scatter:
+          grid_size: !!python/object/apply:numpy.core.multiarray._reconstruct
+            args:
+            - !!python/name:numpy.ndarray ''
+            - !!python/tuple
+              - 0
+            - !!binary |
+              Yg==
+            state: !!python/tuple
+            - 1
+            - !!python/tuple
+              - 3
+            - !!python/object/apply:numpy.dtype
+              args:
+              - i8
+              - false
+              - true
+              state: !!python/tuple
+              - 3
+              - <
+              - null
+              - null
+              - null
+              - -1
+              - -1
+              - 0
+            - false
+            - !!binary |
+              AAEAAAAAAAAAAQAAAAAAAAEAAAAAAAAA
+          num_features: 64
+        voxel_size: &id005
+        - 0.4
+        - 0.4
+        - 4
+      sensor_type: lidar
+    m2:
+      aligner_args:
+        args:
+          dim: 64
+          num_of_blocks: 3
+        core_method: convnext
+        freeze: true
+        spatial_align: false
+      backbone_args:
+        freeze: true
+        inplanes: 128
+        layer_nums:
+        - 3
+        layer_strides:
+        - 2
+        num_filters:
+        - 64
+      camera_mask_args:
+        cav_lidar_range: *id002
+        grid_conf: *id003
+      core_method: lift_splat_shoot
+      encoder_args:
+        anchor_number: 2
+        camera_encoder: EfficientNet
+        data_aug_conf: *id004
+        depth_supervision: true
+        freeze: true
+        grid_conf: *id003
+        img_downsample: 8
+        img_features: 128
+        use_depth_gt: false
+      sensor_type: camera
+    mm_fusion_method: lamma3
+    shrink_header:
+      dim:
+      - 256
+      freeze: true
+      input_dim: 384
+      kernal_size:
+      - 3
+      padding:
+      - 1
+      stride:
+      - 1
+    single_modality: false
+    voxel_size: *id005
+  core_method: point_pillar_lss_lamma2_pyramid_fusion
+name: opv2v_lidarcamera_lamma2_pyramid_fusion
+noise_setting: !!python/object/apply:collections.OrderedDict
+- - - add_noise
+    - false
+optimizer:
+  args:
+    eps: 1.0e-10
+    weight_decay: 0.0001
+  core_method: Adam
+  lr: 0.0002
+postprocess:
+  anchor_args:
+    D: 1
+    H: 256
+    W: 256
+    cav_lidar_range: *id002
+    feature_stride: 2
+    h: 1.56
+    l: 3.9
+    num: 2
+    r: *id006
+    vd: 4
+    vh: 0.4
+    vw: 0.4
+    w: 1.6
+  core_method: VoxelPostprocessor
+  dir_args: *id001
+  gt_range: *id002
+  max_num: 150
+  nms_thresh: 0.15
+  order: hwl
+  target_args:
+    neg_threshold: 0.45
+    pos_threshold: 0.6
+    score_threshold: 0.2
+preprocess:
+  args:
+    max_points_per_voxel: 32
+    max_voxel_test: 70000
+    max_voxel_train: 32000
+    voxel_size: *id005
+  cav_lidar_range: *id002
+  core_method: SpVoxelPreprocessor
+root_dir: data/OPV2V/train
+test_dir: data/OPV2V/test
+train_params:
+  batch_size: 1
+  epoches: 50
+  eval_freq: 2
+  max_cav: 5
+  save_freq: 2
+validate_dir: data/OPV2V/validate
+yaml_parser: load_general_params