feat: add test, docker file and readme

Signed-off-by: Kaan Colak <kaancolak95@gmail.com>

Author: kaancolak

docker/Dockerfile

@@ -1,5 +1,5 @@
-ARG PYTORCH="1.9.0"
-ARG CUDA="11.1"
+ARG PYTORCH="1.13.1"
+ARG CUDA="11.6"
 ARG CUDNN="8"
 
 FROM pytorch/pytorch:${PYTORCH}-cuda${CUDA}-cudnn${CUDNN}-devel
@@ -9,14 +9,6 @@ ENV TORCH_CUDA_ARCH_LIST="6.0 6.1 7.0 7.5 8.0 8.6+PTX" \
     CMAKE_PREFIX_PATH="$(dirname $(which conda))/../" \
     FORCE_CUDA="1"
 
-# Avoid Public GPG key error
-# https://github.com/NVIDIA/nvidia-docker/issues/1631
-RUN rm /etc/apt/sources.list.d/cuda.list \
-    && rm /etc/apt/sources.list.d/nvidia-ml.list \
-    && apt-key del 7fa2af80 \
-    && apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/3bf863cc.pub \
-    && apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64/7fa2af80.pub
-
 # (Optional, use Mirror to speed up downloads)
 # RUN sed -i 's/http:\/\/archive.ubuntu.com\/ubuntu\//http:\/\/mirrors.aliyun.com\/ubuntu\//g' /etc/apt/sources.list && \
 #    pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple
@@ -29,11 +21,11 @@ RUN apt-get update \
 
 # Install MMEngine, MMCV and MMDetection
 RUN pip install openmim && \
-    mim install "mmengine" "mmcv>=2.0.0rc4" "mmdet>=3.0.0"
+    mim install "mmengine" "mmcv>=2.0.0rc4" "mmdet>=3.0.0rc5, <3.3.0"
 
 # Install MMDetection3D
 RUN conda clean --all \
-    && git clone https://github.com/open-mmlab/mmdetection3d.git -b dev-1.x /mmdetection3d \
+    && git clone https://github.com/autowarefoundation/mmdetection3d.git -b main /mmdetection3d \
     && cd /mmdetection3d \
     && pip install --no-cache-dir -e .
 

projects/AutowareCenterPoint/README.md

@@ -0,0 +1,11 @@
+## Introduction
+
+The **[mmdetection3d](https://github.com/open-mmlab/mmdetection3d)** repository includes an additional voxel encoder
+feature for the CenterPoint 3D object detection model, known as voxel center z,
+not originally used in the **[main implementation](https://github.com/tianweiy/CenterPoint)**,
+Autoware maintains consistency with the input size of the original implementation. Consequently,
+to ensure integration with Autoware's lidar centerpoint package, we have forked the original repository and made
+the requisite code modifications.
+
+To train custom CenterPoint models and convert them into ONNX format for deployment in Autoware, please refer to the instructions provided in the README.md file included with
+Autoware's **[lidar_centerpoint](https://autowarefoundation.github.io/autoware.universe/main/perception/lidar_centerpoint/)** package. These instructions will provide a step-by-step guide for training the CenterPoint model.

projects/AutowareCenterPoint/tests/test_pillar_encorder_autoware.py

@@ -0,0 +1,41 @@
+import pytest
+import torch
+
+from mmdet3d.registry import MODELS
+from projects.AutowareCenterPoint.centerpoint.pillar_encoder_autoware import \
+    PillarFeatureNetAutoware # noqa: F401
+
+
+def test_pillar_feature_net_autoware():
+
+    use_voxel_center_z = False
+    if not torch.cuda.is_available():
+        pytest.skip('test requires GPU and torch+cuda')
+    pillar_feature_net_autoware_cfg = dict(
+        type='PillarFeatureNetAutoware',
+        in_channels=4,
+        feat_channels=[64],
+        voxel_size=(0.2, 0.2, 8),
+        point_cloud_range=(-51.2, -51.2, -5.0, 51.2, 51.2, 3.0),
+        norm_cfg=dict(type='BN1d', eps=1e-3, momentum=0.01),
+        use_voxel_center_z=use_voxel_center_z,
+        with_distance=False,
+    )
+    pillar_feature_net_autoware = MODELS.build(pillar_feature_net_autoware_cfg)
+
+    features = torch.rand([97297, 20, 4])
+    num_voxels = torch.randint(1, 100, [97297])
+    coors = torch.randint(0, 100, [97297, 4])
+
+    features = pillar_feature_net_autoware(features, num_voxels, coors)
+
+    if not use_voxel_center_z:
+        assert pillar_feature_net_autoware.pfn_layers[
+            0].linear.in_features == 9
+    else:
+        assert pillar_feature_net_autoware.pfn_layers[
+            0].linear.in_features == 9
+
+    assert pillar_feature_net_autoware.pfn_layers[0].linear.out_features == 64
+
+    assert features.shape == torch.Size([97297, 64])

projects/AutowareCenterPoint/tests/test_t4_dataset.py

@@ -0,0 +1,77 @@
+import numpy as np
+from mmcv.transforms.base import BaseTransform
+from mmengine.registry import TRANSFORMS
+from mmengine.structures import InstanceData
+
+from mmdet3d.structures import Det3DDataSample, LiDARInstance3DBoxes
+from projects.AutowareCenterPoint.datasets.tier4_dataset import T4Dataset
+
+
+def _generate_t4_dataset_config():
+    data_root = 'data/sample_dataset/'
+    ann_file = 'T4Dataset_infos_train.pkl'
+    classes = ['car', 'truck', 'bus', 'bicycle', 'pedestrian']
+
+    if 'Identity' not in TRANSFORMS:
+
+        @TRANSFORMS.register_module()
+        class Identity(BaseTransform):
+
+            def transform(self, info):
+                packed_input = dict(data_samples=Det3DDataSample())
+                if 'ann_info' in info:
+                    packed_input[
+                        'data_samples'].gt_instances_3d = InstanceData()
+                    packed_input[
+                        'data_samples'].gt_instances_3d.labels_3d = info[
+                            'ann_info']['gt_labels_3d']
+                return packed_input
+
+    pipeline = [
+        dict(type='Identity'),
+    ]
+    modality = dict(use_lidar=True, use_camera=True)
+    data_prefix = dict(
+        pts='samples/LIDAR_TOP',
+        img='samples/CAM_BACK_LEFT',
+        sweeps='sweeps/LIDAR_TOP')
+    return data_root, ann_file, classes, data_prefix, pipeline, modality
+
+
+def test_getitem():
+    np.random.seed(0)
+    data_root, ann_file, classes, data_prefix, pipeline, modality = \
+        _generate_t4_dataset_config()
+
+    t4_dataset = T4Dataset(
+        data_root=data_root,
+        ann_file=ann_file,
+        data_prefix=data_prefix,
+        pipeline=pipeline,
+        metainfo=dict(classes=classes),
+        modality=modality)
+
+    t4_dataset.prepare_data(0)
+    input_dict = t4_dataset.get_data_info(0)
+    # assert the the path should contains data_prefix and data_root
+    assert data_prefix['pts'] in input_dict['lidar_points']['lidar_path']
+    assert data_root in input_dict['lidar_points']['lidar_path']
+
+    for cam_id, img_info in input_dict['images'].items():
+        if 'img_path' in img_info:
+            assert data_prefix['img'] in img_info['img_path']
+            assert data_root in img_info['img_path']
+
+    ann_info = t4_dataset.parse_ann_info(input_dict)
+
+    # assert the keys in ann_info and the type
+    assert 'gt_labels_3d' in ann_info
+    assert ann_info['gt_labels_3d'].dtype == np.int64
+    assert len(ann_info['gt_labels_3d']) == 70
+
+    assert 'gt_bboxes_3d' in ann_info
+    assert isinstance(ann_info['gt_bboxes_3d'], LiDARInstance3DBoxes)
+
+    assert len(t4_dataset.metainfo['classes']) == 5
+    assert input_dict['token'] == '5f73a4f0dd74434260bf72821b24c8d4'
+    assert input_dict['timestamp'] == 1697190328.324525