MVX-Net low perfomance pedestrian AOS

### Prerequisite

- [x] I have searched [Issues](https://github.com/open-mmlab/mmdetection3d/issues) and [Discussions](https://github.com/open-mmlab/mmdetection3d/discussions) but cannot get the expected help.
- [x] I have read the [FAQ documentation](https://mmdetection3d.readthedocs.io/en/latest/notes/faq.html) but cannot get the expected help.
- [x] The bug has not been fixed in the [latest version (dev-1.x)](https://github.com/open-mmlab/mmdetection3d/tree/dev-1.x) or [latest version (dev-1.0)](https://github.com/open-mmlab/mmdetection3d/tree/dev-1.0).

### Task

I'm using the official example scripts/configs for the officially supported tasks/models/datasets.

### Branch

main branch https://github.com/open-mmlab/mmdetection3d

### Environment

sys.platform: linux
Python: 3.10.12 (main, Nov  6 2024, 20:22:13) [GCC 11.4.0]
CUDA available: True
MUSA available: False
numpy_random_seed: 2147483648
GPU 0: Orin
CUDA_HOME: /usr/local/cuda
NVCC: Cuda compilation tools, release 12.6, V12.6.77
GCC: aarch64-linux-gnu-gcc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0
PyTorch: 2.5.0
PyTorch compiling details: PyTorch built with:
  - GCC 11.4
  - C++ Version: 201703
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - LAPACK is enabled (usually provided by MKL)
  - NNPACK is enabled
  - CPU capability usage: NO AVX
  - CUDA Runtime 12.6
  - NVCC architecture flags: -gencode;arch=compute_87,code=sm_87
  - CuDNN 90.4
  - Build settings: BLAS_INFO=open, BUILD_TYPE=Release, CUDA_VERSION=12.6, CUDNN_VERSION=9.4.0, CXX_COMPILER=/usr/bin/c++, CXX_FLAGS= -D_GLIBCXX_USE_CXX11_ABI=1 -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -DNDEBUG -DUSE_KINETO -DLIBKINETO_NOROCTRACER -DLIBKINETO_NOXPUPTI=ON -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DSYMBOLICATE_MOBILE_DEBUG_HANDLE -O2 -fPIC -Wall -Wextra -Werror=return-type -Werror=non-virtual-dtor -Werror=range-loop-construct -Werror=bool-operation -Wnarrowing -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-unused-parameter -Wno-strict-overflow -Wno-strict-aliasing -Wno-stringop-overflow -Wsuggest-override -Wno-psabi -Wno-error=old-style-cast -Wno-missing-braces -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, LAPACK_INFO=open, TORCH_VERSION=2.5.0, USE_CUDA=ON, USE_CUDNN=ON, USE_CUSPARSELT=OFF, USE_EIGEN_FOR_BLAS=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_GLOO=1, USE_MKL=OFF, USE_MKLDNN=OFF, USE_MPI=0, USE_NCCL=1, USE_NNPACK=1, USE_OPENMP=ON, USE_ROCM=OFF, USE_ROCM_KERNEL_ASSERT=OFF, 

TorchVision: 0.20.0
OpenCV: 4.10.0-dev
MMEngine: 0.10.5
MMDetection: 3.2.0
MMDetection3D: 1.4.0+962f093
spconv2.0: False

### Reproduces the problem - code sample

# This schedule is mainly used by models with dynamic voxelization
# optimizer
lr = 0.003  # max learning rate
optim_wrapper = dict(
    type='OptimWrapper',
    optimizer=dict(
        type='AdamW', lr=lr, weight_decay=0.001, betas=(0.95, 0.99)),
    clip_grad=dict(max_norm=10, norm_type=2),
)

param_scheduler = [
    dict(type='LinearLR', start_factor=0.1, by_epoch=False, begin=0, end=1000),
    dict(
        type='CosineAnnealingLR',
        begin=0,
        T_max=60,
        end=60,
        by_epoch=True,
        eta_min=1e-5)
]
# training schedule for 1x
train_cfg = dict(type='EpochBasedTrainLoop', max_epochs=60, val_interval=1)
val_cfg = dict(type='ValLoop')
test_cfg = dict(type='TestLoop')

# Default setting for scaling LR automatically
#   - `enable` means enable scaling LR automatically
#       or not by default.
#   - `base_batch_size` = (8 GPUs) x (2 samples per GPU).
auto_scale_lr = dict(enable=False, base_batch_size=2)

_base_ = ['../_base_/schedules/cosine.py', '../_base_/default_runtime.py']

# model settings
voxel_size = [0.05, 0.05, 0.1]
point_cloud_range = [0, -40, -3, 70.4, 40, 1]

model = dict(
    type='DynamicMVXFasterRCNN',
    data_preprocessor=dict(
        type='Det3DDataPreprocessor',
        voxel=True,
        voxel_type='dynamic',
        voxel_layer=dict(
            max_num_points=-1,
            point_cloud_range=point_cloud_range,
            voxel_size=voxel_size,
            max_voxels=(-1, -1)),
        mean=[102.9801, 115.9465, 122.7717],
        std=[1.0, 1.0, 1.0],
        bgr_to_rgb=False,
        pad_size_divisor=32),
    img_backbone=dict(
        type='mmdet.ResNet',
        depth=50,
        num_stages=4,
        out_indices=(0, 1, 2, 3),
        frozen_stages=1,
        norm_cfg=dict(type='BN', requires_grad=False),
        norm_eval=True,
        style='caffe'),
    img_neck=dict(
        type='mmdet.FPN',
        in_channels=[256, 512, 1024, 2048],
        out_channels=256,
        # make the image features more stable numerically to avoid loss nan
        norm_cfg=dict(type='BN', requires_grad=False),
        num_outs=5),
    pts_voxel_encoder=dict(
        type='DynamicVFE',
        in_channels=4,
        feat_channels=[64, 64],
        with_distance=False,
        voxel_size=voxel_size,
        with_cluster_center=True,
        with_voxel_center=True,
        point_cloud_range=point_cloud_range,
        fusion_layer=dict(
            type='PointFusion',
            img_channels=256,
            pts_channels=64,
            mid_channels=128,
            out_channels=128,
            img_levels=[0, 1, 2, 3, 4],
            align_corners=False,
            activate_out=True,
            fuse_out=False)),
    pts_middle_encoder=dict(
        type='SparseEncoder',
        in_channels=128,
        sparse_shape=[41, 1600, 1408],
        order=('conv', 'norm', 'act')),
    pts_backbone=dict(
        type='SECOND',
        in_channels=256,
        layer_nums=[5, 5],
        layer_strides=[1, 2],
        out_channels=[128, 256]),
    pts_neck=dict(
        type='SECONDFPN',
        in_channels=[128, 256],
        upsample_strides=[1, 2],
        out_channels=[256, 256]),
    pts_bbox_head=dict(
        type='Anchor3DHead',
        num_classes=3,
        in_channels=512,
        feat_channels=512,
        use_direction_classifier=True,
        anchor_generator=dict(
            type='Anchor3DRangeGenerator',
            ranges=[
                [0, -40.0, -0.6, 70.4, 40.0, -0.6],
                [0, -40.0, -0.6, 70.4, 40.0, -0.6],
                [0, -40.0, -1.78, 70.4, 40.0, -1.78],
            ],
            sizes=[[0.8, 0.6, 1.73], [1.76, 0.6, 1.73], [3.9, 1.6, 1.56]],
            rotations=[0, 1.57],
            reshape_out=False),
        assigner_per_size=True,
        diff_rad_by_sin=True,
        assign_per_class=True,
        bbox_coder=dict(type='DeltaXYZWLHRBBoxCoder'),
        loss_cls=dict(
            type='mmdet.FocalLoss',
            use_sigmoid=True,
            gamma=2.0,
            alpha=0.25,
            loss_weight=1.0),
        loss_bbox=dict(
            type='mmdet.SmoothL1Loss', beta=1.0 / 9.0, loss_weight=2.0),
        loss_dir=dict(
            type='mmdet.CrossEntropyLoss', use_sigmoid=False,
            loss_weight=0.2)),
    # model training and testing settings
    train_cfg=dict(
        type='EpochBasedTrainLoop',
        pts=dict(
            assigner=[
                dict(  # for Pedestrian
                    type='Max3DIoUAssigner',
                    iou_calculator=dict(type='BboxOverlapsNearest3D'),
                    pos_iou_thr=0.35,
                    neg_iou_thr=0.2,
                    min_pos_iou=0.2,
                    ignore_iof_thr=-1),
                dict(  # for Cyclist
                    type='Max3DIoUAssigner',
                    iou_calculator=dict(type='BboxOverlapsNearest3D'),
                    pos_iou_thr=0.35,
                    neg_iou_thr=0.2,
                    min_pos_iou=0.2,
                    ignore_iof_thr=-1),
                dict(  # for Car
                    type='Max3DIoUAssigner',
                    iou_calculator=dict(type='BboxOverlapsNearest3D'),
                    pos_iou_thr=0.6,
                    neg_iou_thr=0.45,
                    min_pos_iou=0.45,
                    ignore_iof_thr=-1),
            ],
            allowed_border=0,
            pos_weight=-1,
            debug=False)),
    test_cfg=dict(
        pts=dict(
            use_rotate_nms=True,
            nms_across_levels=False,
            nms_thr=0.01,
            score_thr=0.1,
            min_bbox_size=0,
            nms_pre=100,
            max_num=50)))

# dataset settings
dataset_type = 'KittiDataset'
data_root = 'data/kitti/'
class_names = ['Pedestrian', 'Cyclist', 'Car']
metainfo = dict(classes=class_names)
input_modality = dict(use_lidar=True, use_camera=True)
backend_args = None
train_pipeline = [
    dict(
        type='LoadPointsFromFile',
        coord_type='LIDAR',
        load_dim=4,
        use_dim=4,
        backend_args=backend_args),
    dict(type='LoadImageFromFile', backend_args=backend_args),
    dict(type='LoadAnnotations3D', with_bbox_3d=True, with_label_3d=True),
    dict(
        type='RandomResize', scale=[(640, 192), (2560, 768)], keep_ratio=True),
    dict(
        type='GlobalRotScaleTrans',
        rot_range=[-0.78539816, 0.78539816],
        scale_ratio_range=[0.95, 1.05],
        translation_std=[0.2, 0.2, 0.2]),
    dict(type='RandomFlip3D', flip_ratio_bev_horizontal=0.5),
    dict(type='PointsRangeFilter', point_cloud_range=point_cloud_range),
    dict(type='ObjectRangeFilter', point_cloud_range=point_cloud_range),
    dict(type='PointShuffle'),
    dict(
        type='Pack3DDetInputs',
        keys=[
            'points', 'img', 'gt_bboxes_3d', 'gt_labels_3d', 'gt_bboxes',
            'gt_labels'
        ])
]
test_pipeline = [
    dict(
        type='LoadPointsFromFile',
        coord_type='LIDAR',
        load_dim=4,
        use_dim=4,
        backend_args=backend_args),
    dict(type='LoadImageFromFile', backend_args=backend_args),
    dict(
        type='MultiScaleFlipAug3D',
        img_scale=(1280, 384),
        pts_scale_ratio=1,
        flip=False,
        transforms=[
            # Temporary solution, fix this after refactor the augtest
            dict(type='Resize', scale=0, keep_ratio=True),
            dict(
                type='GlobalRotScaleTrans',
                rot_range=[0, 0],
                scale_ratio_range=[1., 1.],
                translation_std=[0, 0, 0]),
            dict(type='RandomFlip3D'),
            dict(
                type='PointsRangeFilter', point_cloud_range=point_cloud_range),
        ]),
    dict(type='Pack3DDetInputs', keys=['points', 'img'])
]
modality = dict(use_lidar=True, use_camera=True)
train_dataloader = dict(
    batch_size=2,
    num_workers=2,
    sampler=dict(type='DefaultSampler', shuffle=True),
    dataset=dict(
        type='RepeatDataset',
        times=2,
        dataset=dict(
            type=dataset_type,
            data_root=data_root,
            modality=modality,
            ann_file='kitti_infos_train.pkl',
            data_prefix=dict(
                pts='training/velodyne_reduced', img='training/image_2'),
            pipeline=train_pipeline,
            filter_empty_gt=False,
            metainfo=metainfo,
            # we use box_type_3d='LiDAR' in kitti and nuscenes dataset
            # and box_type_3d='Depth' in sunrgbd and scannet dataset.
            box_type_3d='LiDAR',
            backend_args=backend_args)))

val_dataloader = dict(
    batch_size=1,
    num_workers=1,
    sampler=dict(type='DefaultSampler', shuffle=False),
    dataset=dict(
        type=dataset_type,
        data_root=data_root,
        modality=modality,
        ann_file='kitti_infos_val.pkl',
        data_prefix=dict(
            pts='training/velodyne_reduced', img='training/image_2'),
        pipeline=test_pipeline,
        metainfo=metainfo,
        test_mode=True,
        box_type_3d='LiDAR',
        backend_args=backend_args))
test_dataloader = dict(
    batch_size=1,
    num_workers=1,
    sampler=dict(type='DefaultSampler', shuffle=False),
    dataset=dict(
        type=dataset_type,
        data_root=data_root,
        ann_file='kitti_infos_val.pkl',
        modality=modality,
        data_prefix=dict(
            pts='training/velodyne_reduced', img='training/image_2'),
        pipeline=test_pipeline,
        metainfo=metainfo,
        test_mode=True,
        box_type_3d='LiDAR',
        backend_args=backend_args))
# optim_wrapper se sobreescribe en cosine.py 
optim_wrapper = dict(
    optimizer=dict(weight_decay=0.01),
    clip_grad=dict(max_norm=35, norm_type=2),
)
val_evaluator = dict(
    type='KittiMetric', ann_file='data/kitti/kitti_infos_val.pkl')
test_evaluator = val_evaluator

# You may need to download the model first is the network is unstable
load_from = 'https://download.openmmlab.com/mmdetection3d/pretrain_models/mvx_faster_rcnn_detectron2-caffe_20e_coco-pretrain_gt-sample_kitti-3-class_moderate-79.3_20200207-a4a6a3c7.pth'  # noqa


### Reproduces the problem - command or script

python3 tools/train.py work_dirs/mvxnet_fpn_dv_second_secfpn_8xb2-80e_kitti-3d-3class/mvxnet_fpn_dv_second_secfpn_8xb2-80e_kitti-3d-3class.py 

### Reproduces the problem - error message

**EPOCH 40**

![Image](https://github.com/user-attachments/assets/074be70f-1e4f-4c02-8582-896e75a2a023)

**EPOCH 49**

![Image](https://github.com/user-attachments/assets/7ff0d16e-5969-4257-b7d0-5760b133844e)

### Additional information

First, I trained MVX-Net for 40 epochs using the original config file (I only changed the batch size to 2 for 1 GPU). The Pedestrian AOS AP results for the 40th epoch were quite low, less than 50%, which I think is very poor. I tried resuming the training for 10 more epochs to improve the results, but it didn't seem to get better (on the contrary, they've worsen) .

I'm using the KITTI dataset with 3769 training and 3712 validation samples. I got the ImageSets from this model: [VirConv ImageSets](https://github.com/hailanyi/VirConv/tree/master/data/kitti/ImageSets)

I would like to know if anyone has achieved better AOS results for pedestrians with MVX-Net and, if so, what changes I should make to the config file. I don't have much experience with model training, so I welcome any advice and support. Thank you!

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

MVX-Net low perfomance pedestrian AOS #3080

Prerequisite

Task

Branch

Environment

Reproduces the problem - code sample

This schedule is mainly used by models with dynamic voxelization

optimizer

training schedule for 1x

Default setting for scaling LR automatically

- `enable` means enable scaling LR automatically

or not by default.

- `base_batch_size` = (8 GPUs) x (2 samples per GPU).

model settings

dataset settings

optim_wrapper se sobreescribe en cosine.py

You may need to download the model first is the network is unstable

Reproduces the problem - command or script

Reproduces the problem - error message

Additional information

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

MVX-Net low perfomance pedestrian AOS #3080

Description

Prerequisite

Task

Branch

Environment

Reproduces the problem - code sample

This schedule is mainly used by models with dynamic voxelization

optimizer

training schedule for 1x

Default setting for scaling LR automatically

- enable means enable scaling LR automatically

or not by default.

- base_batch_size = (8 GPUs) x (2 samples per GPU).

model settings

dataset settings

optim_wrapper se sobreescribe en cosine.py

You may need to download the model first is the network is unstable

Reproduces the problem - command or script

Reproduces the problem - error message

Additional information

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

- `enable` means enable scaling LR automatically

- `base_batch_size` = (8 GPUs) x (2 samples per GPU).