Bump version to V1.0.0rc0

Bump version to V1.0.0rc0

configs/h3dnet/README.md

 # H3DNet: 3D Object Detection Using Hybrid Geometric Primitives
 
-## Introduction
+> [H3DNet: 3D Object Detection Using Hybrid Geometric Primitives](https://arxiv.org/abs/2006.05682)
 
 <!-- [ALGORITHM] -->
 
-We implement H3DNet and provide the result and checkpoints on ScanNet datasets.
+## Abstract
 
-```
-@inproceedings{zhang2020h3dnet,
-    author = {Zhang, Zaiwei and Sun, Bo and Yang, Haitao and Huang, Qixing},
-    title = {H3DNet: 3D Object Detection Using Hybrid Geometric Primitives},
-    booktitle = {Proceedings of the European Conference on Computer Vision},
-    year = {2020}
-}
-```
+We introduce H3DNet, which takes a colorless 3D point cloud as input and outputs a collection of oriented object bounding boxes (or BB) and their semantic labels. The critical idea of H3DNet is to predict a hybrid set of geometric primitives, i.e., BB centers, BB face centers, and BB edge centers. We show how to convert the predicted geometric primitives into object proposals by defining a distance function between an object and the geometric primitives. This distance function enables continuous optimization of object proposals, and its local minimums provide high-fidelity object proposals. H3DNet then utilizes a matching and refinement module to classify object proposals into detected objects and fine-tune the geometric parameters of the detected objects. The hybrid set of geometric primitives not only provides more accurate signals for object detection than using a single type of geometric primitives, but it also provides an overcomplete set of constraints on the resulting 3D layout. Therefore, H3DNet can tolerate outliers in predicted geometric primitives. Our model achieves state-of-the-art 3D detection results on two large datasets with real 3D scans, ScanNet and SUN RGB-D.
+
+<div align=center>
+<img src="https://user-images.githubusercontent.com/36950400/143868884-26f7fc63-93fd-48cb-a469-e2f55fda5550.png" width="800"/>
+</div>
+
+## Introduction
+
+We implement H3DNet and provide the result and checkpoints on ScanNet datasets.
 
-## Results
+## Results and models
 
 ### ScanNet
 
@@ -30,3 +31,14 @@ python ./tools/model_converters/convert_h3dnet_checkpoints.py ${ORIGINAL_CHECKPO
 ```
 
 Then you can use the converted checkpoints following [getting_started.md](../../docs/en/getting_started.md).
+
+## Citation
+
+```latex
+@inproceedings{zhang2020h3dnet,
+    author = {Zhang, Zaiwei and Sun, Bo and Yang, Haitao and Huang, Qixing},
+    title = {H3DNet: 3D Object Detection Using Hybrid Geometric Primitives},
+    booktitle = {Proceedings of the European Conference on Computer Vision},
+    year = {2020}
+}
+```

configs/imvotenet/README.md

 # ImVoteNet: Boosting 3D Object Detection in Point Clouds with Image Votes
 
-## Introduction
+> [ImVoteNet: Boosting 3D Object Detection in Point Clouds with Image Votes](https://arxiv.org/abs/2001.10692)
 
 <!-- [ALGORITHM] -->
 
-We implement ImVoteNet and provide the result and checkpoints on SUNRGBD.
+## Abstract
 
-```
-@inproceedings{qi2020imvotenet,
-  title={Imvotenet: Boosting 3D object detection in point clouds with image votes},
-  author={Qi, Charles R and Chen, Xinlei and Litany, Or and Guibas, Leonidas J},
-  booktitle={Proceedings of the IEEE/CVF conference on computer vision and pattern recognition},
-  pages={4404--4413},
-  year={2020}
-}
-```
+3D object detection has seen quick progress thanks to advances in deep learning on point clouds. A few recent works have even shown state-of-the-art performance with just point clouds input (e.g. VOTENET). However, point cloud data have inherent limitations. They are sparse, lack color information and often suffer from sensor noise. Images, on the other hand, have high resolution and rich texture. Thus they can complement the 3D geometry provided by point clouds. Yet how to effectively use image information to assist point cloud based detection is still an open question. In this work, we build on top of VOTENET and propose a 3D detection architecture called IMVOTENET specialized for RGB-D scenes. IMVOTENET is based on fusing 2D votes in images and 3D votes in point clouds. Compared to prior work on multi-modal detection, we explicitly extract both geometric and semantic features from the 2D images. We leverage camera parameters to lift these features to 3D. To improve the synergy of 2D-3D feature fusion, we also propose a multi-tower training scheme. We validate our model on the challenging SUN RGB-D dataset, advancing state-of-the-art results by 5.7 mAP. We also provide rich ablation studies to analyze the contribution of each design choice.
+
+<div align=center>
+<img src="https://user-images.githubusercontent.com/36950400/143869878-a2ae7f43-55c3-4b95-af09-8f97dfd975f4.png" width="800"/>
+</div>
+
+## Introduction
+
+We implement ImVoteNet and provide the result and checkpoints on SUNRGBD.
 
-## Results
+## Results and models
 
 ### SUNRGBD-2D (Stage 1, image branch pre-train)
 
@@ -29,3 +29,15 @@ We implement ImVoteNet and provide the result and checkpoints on SUNRGBD.
 |  Backbone   | Lr schd | Mem (GB) | Inf time (fps) | AP@0.25 |AP@0.5| Download |
 | :---------: | :-----: | :------: | :------------: | :----: |:----: | :------: |
 |    [PointNet++](./imvotenet_stage2_16x8_sunrgbd-3d-10class.py)     |  3x    |9.4| |64.04||[model](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/imvotenet/imvotenet_stage2_16x8_sunrgbd-3d-10class/imvotenet_stage2_16x8_sunrgbd-3d-10class_20210323_184021-d44dcb66.pth) &#124; [log](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/imvotenet/imvotenet_stage2_16x8_sunrgbd-3d-10class/imvotenet_stage2_16x8_sunrgbd-3d-10class_20210323_184021.log.json)|
+
+## Citation
+
+```latex
+@inproceedings{qi2020imvotenet,
+  title={Imvotenet: Boosting 3D object detection in point clouds with image votes},
+  author={Qi, Charles R and Chen, Xinlei and Litany, Or and Guibas, Leonidas J},
+  booktitle={Proceedings of the IEEE/CVF conference on computer vision and pattern recognition},
+  pages={4404--4413},
+  year={2020}
+}
+```

configs/imvoxelnet/README.md

 # ImVoxelNet: Image to Voxels Projection for Monocular and Multi-View General-Purpose 3D Object Detection
 
-## Introduction
+> [ImVoxelNet: Image to Voxels Projection for Monocular and Multi-View General-Purpose 3D Object Detection](https://arxiv.org/abs/2106.01178)
 
 <!-- [ALGORITHM] -->
 
+## Abstract
+
+In this paper, we introduce the task of multi-view RGB-based 3D object detection as an end-to-end optimization problem. To address this problem, we propose ImVoxelNet, a novel fully convolutional method of 3D object detection based on posed monocular or multi-view RGB images. The number of monocular images in each multiview input can variate during training and inference; actually, this number might be unique for each multi-view input. ImVoxelNet successfully handles both indoor and outdoor scenes, which makes it general-purpose. Specifically, it achieves state-of-the-art results in car detection on KITTI (monocular) and nuScenes (multi-view) benchmarks among all methods that accept RGB images. Moreover, it surpasses existing RGB-based 3D object detection methods on the SUN RGB-D dataset. On ScanNet, ImVoxelNet sets a new benchmark for multi-view 3D object detection.
+
+<div align=center>
+<img src="https://user-images.githubusercontent.com/36950400/143871445-38a55168-b8cd-4520-8ed6-f5c8c8ea304a.png" width="800"/>
+</div>
+
+## Introduction
+
 We implement a monocular 3D detector ImVoxelNet and provide its results and checkpoints on KITTI dataset.
 Results for SUN RGB-D, ScanNet and nuScenes are currently available in ImVoxelNet authors
 [repo](https://github.com/saic-vul/imvoxelnet) (based on mmdetection3d).
 
-```
-@inproceedings{rukhovich2022imvoxelnet,
-  title={Imvoxelnet: Image to voxels projection for monocular and multi-view general-purpose 3d object detection},
-  author={Rukhovich, Danila and Vorontsova, Anna and Konushin, Anton},
-  booktitle={Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision},
-  pages={2397--2406},
-  year={2022}
-}
-```
-
-## Results
+## Results and models
 
 ### KITTI
 
 |  Backbone   |Class| Lr schd | Mem (GB) | Inf time (fps) | mAP | Download |
 | :---------: | :-----: |:-----: | :------: | :------------: | :----: |:----: |
 | [ResNet-50](./imvoxelnet_kitti-3d-car.py) | Car | 3x | | |17.4|[model](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/imvoxelnet/imvoxelnet_kitti-3d-car_20210610_152323-b9abba85.pth) &#124; [log](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/imvoxelnet/imvoxelnet_kitti-3d-car_20210610_152323.log.json)|
+
+## Citation
+
+```latex
+@article{rukhovich2021imvoxelnet,
+  title={ImVoxelNet: Image to Voxels Projection for Monocular and Multi-View General-Purpose 3D Object Detection},
+  author={Danila Rukhovich, Anna Vorontsova, Anton Konushin},
+  journal={arXiv preprint arXiv:2106.01178},
+  year={2021}
+}
+```

configs/imvoxelnet/imvoxelnet_4x8_kitti-3d-car.py

@@ -25,7 +25,7 @@ model = dict(
         anchor_generator=dict(
             type='AlignedAnchor3DRangeGenerator',
             ranges=[[-0.16, -39.68, -1.78, 68.96, 39.68, -1.78]],
-            sizes=[[1.6, 3.9, 1.56]],
+            sizes=[[3.9, 1.6, 1.56]],
             rotations=[0, 1.57],
             reshape_out=True),
         diff_rad_by_sin=True,

configs/monoflex/README.md

+# Objects are Different: Flexible Monocular 3D Object Detection
+
+> [Objects are Different: Flexible Monocular 3D Object Detection](https://arxiv.org/abs/2104.02323)
+
+<!-- [ALGORITHM] -->
+
+## Abstract
+
+The precise localization of 3D objects from a single image without depth information is a highly challenging problem. Most existing methods adopt the same approach for all objects regardless of their diverse distributions, leading to limited performance for truncated objects. In this paper, we propose a flexible framework for monocular 3D object detection which explicitly decouples the truncated objects and adaptively combines multiple approaches for object depth estimation. Specifically, we decouple the edge of the feature map for predicting long-tail truncated objects so that the optimization of normal objects is not influenced. Furthermore, we formulate the object depth estimation as an uncertainty-guided ensemble of directly regressed object depth and solved depths from different groups of keypoints. Experiments demonstrate that our method outperforms the state-of-the-art method by relatively 27% for the moderate level and 30% for the hard level in the test set of KITTI benchmark while maintaining real-time efficiency.
+
+<div align=center>
+<img src="https://user-images.githubusercontent.com/36950400/153138824-d54a7a47-773f-42f9-8a51-b0a71078593e.png" width="800"/>
+</div>
+
+## Introduction
+
+We implement MonoFlex and provide the results and checkpoints on KITTI dataset.
+
+## Results and models
+
+### KITTI
+
+|  Backbone   | Lr schd | Mem (GB) | Inf time (fps) | mAP | Download |
+| :---------: | :-----: | :------: | :------------: | :----: | :------: |
+|[DLA34](./monoflex_dla34_pytorch_dlaneck_gn-all_2x4_6x_kitti-mono3d.py)|6x|9.64||21.86|[model](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/monoflex/monoflex_dla34_pytorch_dlaneck_gn-all_2x4_6x_kitti-mono3d_20211228_027553-d46d9bb0.pth) &#124; [log](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/monoflex/monoflex_dla34_pytorch_dlaneck_gn-all_2x4_6x_kitti-mono3d_20211228_027553.log.json)
+
+Note: mAP represents Car moderate 3D strict AP11 results.
+Detailed performance on KITTI 3D detection (3D/BEV) is as follows, evaluated by AP11 and AP40 metric:
+
+|             |     Easy      |    Moderate    |     Hard      |
+|-------------|:-------------:|:--------------:|:-------------:|
+| Car (AP11)  | 28.02 / 36.11 | 21.86 / 29.46  | 19.01 / 24.83 |
+| Car (AP40)  | 23.22 / 32.74 | 17.18 / 24.02  | 15.13 / 20.67 |
+
+Note: mAP represents Car moderate 3D strict AP11 / AP40 results. Because of the limited data for pedestrians and cyclists, the detection performance for these two classes is usually unstable. Therefore, we only list car detection results here. In addition, the AP11 result may fluctuate in a larger range (~1 AP), so AP40 is a more recommended metric for reference due to its much better stability.
+
+## Citation
+
+```latex
+@InProceedings{MonoFlex,
+    author    = {Zhang, Yunpeng and Lu, Jiwen and Zhou, Jie},
+    title     = {Objects Are Different: Flexible Monocular 3D Object Detection},
+    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
+    month     = {June},
+    year      = {2021},
+    pages     = {3289-3298}
+}
+```

configs/monoflex/metafile.yml

+Collections:
+  - Name: MonoFlex
+    Metadata:
+      Training Data: KITTI
+      Training Techniques:
+        - Adam
+      Training Resources: 2x V100 GPUS
+      Architecture:
+        - MonoFlexHead
+        - DLA
+    Paper:
+      URL: https://arxiv.org/abs/2104.02323
+      Title: 'Objects are Different: Flexible Monocular 3D Object Detection'
+    README: configs/monoflex/README.md
+    Code:
+      URL: https://github.com/open-mmlab/mmdetection3d/blob/v1.0.0.dev0/mmdet3d/models/detectors/monoflex.py#L7
+      Version: v1.0.0
+
+Models:
+  - Name: monoflex_dla34_pytorch_dlaneck_gn-all_2x4_6x_kitti-mono3d
+    In Collection: MonoFlex
+    Config: configs/monoflex/monoflex_dla34_pytorch_dlaneck_gn-all_2x4_6x_kitti-mono3d.py
+    Metadata:
+      Training Memory (GB): 9.64
+    Results:
+      - Task: 3D Object Detection
+        Dataset: KITTI
+        Metrics:
+          mAP: 21.98
+    Weights: https://download.openmmlab.com/mmdetection3d/v0.1.0_models/monoflex/monoflex_dla34_pytorch_dlaneck_gn-all_2x4_6x_kitti-mono3d_20211228_027553-d46d9bb0.pth

configs/mvxnet/README.md

 # MVX-Net: Multimodal VoxelNet for 3D Object Detection
 
-## Introduction
+> [MVX-Net: Multimodal VoxelNet for 3D Object Detection](https://arxiv.org/abs/1904.01649)
 
 <!-- [ALGORITHM] -->
 
+## Abstract
+
+Many recent works on 3D object detection have focused on designing neural network architectures that can consume point cloud data. While these approaches demonstrate encouraging performance, they are typically based on a single modality and are unable to leverage information from other modalities, such as a camera. Although a few approaches fuse data from different modalities, these methods either use a complicated pipeline to process the modalities sequentially, or perform late-fusion and are unable to learn interaction between different modalities at early stages. In this work, we present PointFusion and VoxelFusion: two simple yet effective early-fusion approaches to combine the RGB and point cloud modalities, by leveraging the recently introduced VoxelNet architecture. Evaluation on the KITTI dataset demonstrates significant improvements in performance over approaches which only use point cloud data. Furthermore, the proposed method provides results competitive with the state-of-the-art multimodal algorithms, achieving top-2 ranking in five of the six bird's eye view and 3D detection categories on the KITTI benchmark, by using a simple single stage network.
+
+<div align=center>
+<img src="https://user-images.githubusercontent.com/79644370/143880819-560675ca-e7e3-4d77-8808-ea661ff8e6e6.png" width="800"/>
+</div>
+
+## Introduction
+
 We implement MVX-Net and provide its results and models on KITTI dataset.
 
-```
+## Results and models
+
+### KITTI
+
+|  Backbone   |Class| Lr schd | Mem (GB) | Inf time (fps) | mAP | Download |
+| :---------: | :-----: | :------: | :------------: | :----: |:----: | :------: |
+|    [SECFPN](./dv_mvx-fpn_second_secfpn_adamw_2x8_80e_kitti-3d-3class.py)|3 Class|cosine 80e|6.7||63.0|[model](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/mvxnet/dv_mvx-fpn_second_secfpn_adamw_2x8_80e_kitti-3d-3class/dv_mvx-fpn_second_secfpn_adamw_2x8_80e_kitti-3d-3class_20200621_003904-10140f2d.pth) &#124; [log](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/mvxnet/dv_mvx-fpn_second_secfpn_adamw_2x8_80e_kitti-3d-3class/dv_mvx-fpn_second_secfpn_adamw_2x8_80e_kitti-3d-3class_20200621_003904.log.json)|
+
+## Citation
+
+```latex
 @inproceedings{sindagi2019mvx,
   title={MVX-Net: Multimodal voxelnet for 3D object detection},
   author={Sindagi, Vishwanath A and Zhou, Yin and Tuzel, Oncel},
@@ -15,13 +35,4 @@ We implement MVX-Net and provide its results and models on KITTI dataset.
   year={2019},
   organization={IEEE}
 }
-
 ```
-
-## Results
-
-### KITTI
-
-|  Backbone   |Class| Lr schd | Mem (GB) | Inf time (fps) | mAP | Download |
-| :---------: | :-----: | :------: | :------------: | :----: |:----: | :------: |
-|    [SECFPN](./dv_mvx-fpn_second_secfpn_adamw_2x8_80e_kitti-3d-3class.py)|3 Class|cosine 80e|6.7||63.0|[model](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/mvxnet/dv_mvx-fpn_second_secfpn_adamw_2x8_80e_kitti-3d-3class/dv_mvx-fpn_second_secfpn_adamw_2x8_80e_kitti-3d-3class_20200621_003904-10140f2d.pth) &#124; [log](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/mvxnet/dv_mvx-fpn_second_secfpn_adamw_2x8_80e_kitti-3d-3class/dv_mvx-fpn_second_secfpn_adamw_2x8_80e_kitti-3d-3class_20200621_003904.log.json)|

configs/mvxnet/dv_mvx-fpn_second_secfpn_adamw_2x8_80e_kitti-3d-3class.py

@@ -74,7 +74,7 @@ model = dict(
                 [0, -40.0, -0.6, 70.4, 40.0, -0.6],
                 [0, -40.0, -1.78, 70.4, 40.0, -1.78],
             ],
-            sizes=[[0.6, 0.8, 1.73], [0.6, 1.76, 1.73], [1.6, 3.9, 1.56]],
+            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,

configs/nuimages/README.md

 # NuImages Results
 
-## Introduction
-
 <!-- [DATASET] -->
 
+## Introduction
+
 We support and provide some baseline results on [nuImages dataset](https://www.nuscenes.org/nuimages).
 We follow the class mapping in nuScenes dataset, which maps the original categories into 10 foreground categories.
 The convert script can be found [here](https://github.com/open-mmlab/mmdetection3d/blob/master/tools/data_converter/nuimage_converter.py).
@@ -15,7 +15,7 @@ We will support panoptic segmentation models in the future.
 The dataset converted by the script of v0.6.0 only supports instance segmentation. Since v0.7.0, we also support to produce semantic segmentation mask of each image; thus, we can train HTC or semantic segmentation models using the dataset. To convert the nuImages dataset into COCO format, please use the command below:
 
 ```shell
-python -u tools/data_converter/nuimage_converter.py --data-root ${DATA_ROOT} --version ${VERIONS} \
+python -u tools/data_converter/nuimage_converter.py --data-root ${DATA_ROOT} --version ${VERSIONS} \
                                                     --out-dir ${OUT_DIR} --nproc ${NUM_WORKERS} --extra-tag ${TAG}
 ```
 
@@ -25,7 +25,7 @@ python -u tools/data_converter/nuimage_converter.py --data-root ${DATA_ROOT} --v
 - `--nproc`: number of workers for data preparation, defaults to `4`. Larger number could reduce the preparation time as images are processed in parallel.
 - `--extra-tag`: extra tag of the annotations, defaults to `nuimages`. This can be used to separate different annotations processed in different time for study.
 
-## Results
+## Results and models
 
 ### Instance Segmentation
 
@@ -55,4 +55,4 @@ We report Mask R-CNN and Cascade Mask R-CNN results on nuimages.
 1. `IN` means only using ImageNet pre-trained backbone. `IN+COCO-Nx` and `IN+COCO-Ne` means the backbone is first pre-trained on ImageNet, and then the detector is pre-trained on COCO train2017 dataset by `Nx` and `N` epochs schedules, respectively.
 2. All the training hyper-parameters follow the standard schedules on COCO dataset except that the images are resized from
 1280 x 720 to 1920 x 1080 (relative ratio 0.8 to 1.2) since the images are in size 1600 x 900.
-3. The class order in the detectors released in v0.6.0 is different from the order in the configs because the bug in the convertion script. This bug has been fixed since v0.7.0 and the models trained by the correct class order are also released. If you used nuImages since v0.6.0, please re-convert the data through the convertion script using the above-mentioned command.
+3. The class order in the detectors released in v0.6.0 is different from the order in the configs because the bug in the conversion script. This bug has been fixed since v0.7.0 and the models trained by the correct class order are also released. If you used nuImages since v0.6.0, please re-convert the data through the conversion script using the above-mentioned command.

configs/nuimages/metafile.yml

-Collections:
-  - Name: Mask R-CNN
-    Metadata:
-      Training Data: nuImages
-      Training Techniques:
-        - SGD with Momentum
-      Architecture:
-        - RoI Align
-        - RPN
-    README: configs/nuimages/README.md
-    Code:
-      Version: v0.6.0
-
 Models:
   - Name: mask_rcnn_r50_fpn_1x_nuim
     In Collection: Mask R-CNN

configs/paconv/README.md

 # PAConv: Position Adaptive Convolution with Dynamic Kernel Assembling on Point Clouds
 
-## Introduction
+> [PAConv: Position Adaptive Convolution with Dynamic Kernel Assembling on Point Clouds](https://arxiv.org/abs/2103.14635)
 
 <!-- [ALGORITHM] -->
 
-We implement PAConv and provide the result and checkpoints on S3DIS dataset.
+## Abstract
 
-```
-@inproceedings{xu2021paconv,
-  title={PAConv: Position Adaptive Convolution with Dynamic Kernel Assembling on Point Clouds},
-  author={Xu, Mutian and Ding, Runyu and Zhao, Hengshuang and Qi, Xiaojuan},
-  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
-  pages={3173--3182},
-  year={2021}
-}
-```
+We introduce Position Adaptive Convolution (PAConv), a generic convolution operation for 3D point cloud processing. The key of PAConv is to construct the convolution kernel by dynamically assembling basic weight matrices stored in Weight Bank, where the coefficients of these weight matrices are self-adaptively learned from point positions through ScoreNet. In this way, the kernel is built in a data-driven manner, endowing PAConv with more flexibility than 2D convolutions to better handle the irregular and unordered point cloud data. Besides, the complexity of the learning process is reduced by combining weight matrices instead of brutally predicting kernels from point positions.
+Furthermore, different from the existing point convolution operators whose network architectures are often heavily engineered, we integrate our PAConv into classical MLP-based point cloud pipelines without changing network configurations. Even built on simple networks, our method still approaches or even surpasses the state-of-the-art models, and significantly improves baseline performance on both classification and segmentation tasks, yet with decent efficiency. Thorough ablation studies and visualizations are provided to understand PAConv.
+
+<div align=center>
+<img src="https://user-images.githubusercontent.com/79644370/143881915-003d5f10-3999-474e-969a-c354cb738a11.png" width="800"/>
+</div>
+
+## Introduction
+
+We implement PAConv and provide the result and checkpoints on S3DIS dataset.
 
 **Notice**: The original PAConv paper used step learning rate schedule. We discovered that cosine schedule achieves slightly better results and adopt it in our implementations.
 
-## Results
+## Results and models
 
 ### S3DIS
 
@@ -36,3 +37,15 @@ We implement PAConv and provide the result and checkpoints on S3DIS dataset.
 ## Indeterminism
 
 Since PAConv testing adopts sliding patch inference which involves random point sampling, and the test script uses fixed random seeds while the random seeds of validation in training are not fixed, the test results may be slightly different from the results reported above.
+
+## Citation
+
+```latex
+@inproceedings{xu2021paconv,
+  title={PAConv: Position Adaptive Convolution with Dynamic Kernel Assembling on Point Clouds},
+  author={Xu, Mutian and Ding, Runyu and Zhao, Hengshuang and Qi, Xiaojuan},
+  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
+  pages={3173--3182},
+  year={2021}
+}
+```

configs/parta2/README.md

 # From Points to Parts: 3D Object Detection from Point Cloud with Part-aware and Part-aggregation Network
 
-## Introduction
+> [From Points to Parts: 3D Object Detection from Point Cloud with Part-aware and Part-aggregation Network](https://arxiv.org/abs/1907.03670)
 
 <!-- [ALGORITHM] -->
 
+## Abstract
+
+3D object detection from LiDAR point cloud is a challenging problem in 3D scene understanding and has many practical applications. In this paper, we extend our preliminary work PointRCNN to a novel and strong point-cloud-based 3D object detection framework, the part-aware and aggregation neural network (Part-A2 net). The whole framework consists of the part-aware stage and the part-aggregation stage. Firstly, the part-aware stage for the first time fully utilizes free-of-charge part supervisions derived from 3D ground-truth boxes to simultaneously predict high quality 3D proposals and accurate intra-object part locations. The predicted intra-object part locations within the same proposal are grouped by our new-designed RoI-aware point cloud pooling module, which results in an effective representation to encode the geometry-specific features of each 3D proposal. Then the part-aggregation stage learns to re-score the box and refine the box location by exploring the spatial relationship of the pooled intra-object part locations. Extensive experiments are conducted to demonstrate the performance improvements from each component of our proposed framework. Our Part-A2 net outperforms all existing 3D detection methods and achieves new state-of-the-art on KITTI 3D object detection dataset by utilizing only the LiDAR point cloud data.
+
+<div align=center>
+<img src="https://user-images.githubusercontent.com/79644370/143882774-6fc5f736-10d1-499a-8929-ca0768419049.png" width="800"/>
+</div>
+
+## Introduction
+
 We implement Part-A^2 and provide its results and checkpoints on KITTI dataset.
 
-```
+## Results and models
+
+### KITTI
+
+|  Backbone   |Class| Lr schd | Mem (GB) | Inf time (fps) | mAP | Download |
+| :---------: | :-----: |:-----: | :------: | :------------: | :----: |:----: |
+|    [SECFPN](./hv_PartA2_secfpn_2x8_cyclic_80e_kitti-3d-3class.py) |3 Class|cyclic 80e|4.1||67.9|[model](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/parta2/hv_PartA2_secfpn_2x8_cyclic_80e_kitti-3d-3class/hv_PartA2_secfpn_2x8_cyclic_80e_kitti-3d-3class_20200620_230724-a2672098.pth) &#124; [log](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/parta2/hv_PartA2_secfpn_2x8_cyclic_80e_kitti-3d-3class/hv_PartA2_secfpn_2x8_cyclic_80e_kitti-3d-3class_20200620_230724.log.json)|
+|    [SECFPN](./hv_PartA2_secfpn_2x8_cyclic_80e_kitti-3d-car.py) |Car |cyclic 80e|4.0||79.16|[model](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/parta2/hv_PartA2_secfpn_2x8_cyclic_80e_kitti-3d-car/hv_PartA2_secfpn_2x8_cyclic_80e_kitti-3d-car_20200620_230755-f2a38b9a.pth) &#124; [log](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/parta2/hv_PartA2_secfpn_2x8_cyclic_80e_kitti-3d-car/hv_PartA2_secfpn_2x8_cyclic_80e_kitti-3d-car_20200620_230755.log.json)|
+
+## Citation
+
+```latex
 @article{shi2020points,
   title={From points to parts: 3d object detection from point cloud with part-aware and part-aggregation network},
   author={Shi, Shaoshuai and Wang, Zhe and Shi, Jianping and Wang, Xiaogang and Li, Hongsheng},
@@ -15,12 +36,3 @@ We implement Part-A^2 and provide its results and checkpoints on KITTI dataset.
   publisher={IEEE}
 }
 ```
-
-## Results
-
-### KITTI
-
-|  Backbone   |Class| Lr schd | Mem (GB) | Inf time (fps) | mAP | Download |
-| :---------: | :-----: |:-----: | :------: | :------------: | :----: |:----: |
-|    [SECFPN](./hv_PartA2_secfpn_2x8_cyclic_80e_kitti-3d-3class.py) |3 Class|cyclic 80e|4.1||67.9|[model](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/parta2/hv_PartA2_secfpn_2x8_cyclic_80e_kitti-3d-3class/hv_PartA2_secfpn_2x8_cyclic_80e_kitti-3d-3class_20200620_230724-a2672098.pth) &#124; [log](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/parta2/hv_PartA2_secfpn_2x8_cyclic_80e_kitti-3d-3class/hv_PartA2_secfpn_2x8_cyclic_80e_kitti-3d-3class_20200620_230724.log.json)|
-|    [SECFPN](./hv_PartA2_secfpn_2x8_cyclic_80e_kitti-3d-car.py) |Car |cyclic 80e|4.0||79.16|[model](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/parta2/hv_PartA2_secfpn_2x8_cyclic_80e_kitti-3d-car/hv_PartA2_secfpn_2x8_cyclic_80e_kitti-3d-car_20200620_230755-f2a38b9a.pth) &#124; [log](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/parta2/hv_PartA2_secfpn_2x8_cyclic_80e_kitti-3d-car/hv_PartA2_secfpn_2x8_cyclic_80e_kitti-3d-car_20200620_230755.log.json)|

configs/parta2/hv_PartA2_secfpn_2x8_cyclic_80e_kitti-3d-car.py

@@ -10,7 +10,7 @@ model = dict(
             _delete_=True,
             type='Anchor3DRangeGenerator',
             ranges=[[0, -40.0, -1.78, 70.4, 40.0, -1.78]],
-            sizes=[[1.6, 3.9, 1.56]],
+            sizes=[[3.9, 1.6, 1.56]],
             rotations=[0, 1.57],
             reshape_out=False)),
     roi_head=dict(

configs/pgd/README.md

+# Probabilistic and Geometric Depth: Detecting Objects in Perspective
+
+> [Probabilistic and Geometric Depth: Detecting Objects in Perspective](https://arxiv.org/abs/2107.14160)
+
+<!-- [ALGORITHM] -->
+
+## Abstract
+
+3D object detection is an important capability needed in various practical applications such as driver assistance systems. Monocular 3D detection, as a representative general setting among image-based approaches, provides a more economical solution than conventional settings relying on LiDARs but still yields unsatisfactory results. This paper first presents a systematic study on this problem. We observe that the current monocular 3D detection can be simplified as an instance depth estimation problem: The inaccurate instance depth blocks all the other 3D attribute predictions from improving the overall detection performance. Moreover, recent methods directly estimate the depth based on isolated instances or pixels while ignoring the geometric relations across different objects. To this end, we construct geometric relation graphs across predicted objects and use the graph to facilitate depth estimation. As the preliminary depth estimation of each instance is usually inaccurate in this ill-posed setting, we incorporate a probabilistic representation to capture the uncertainty. It provides an important indicator to identify confident predictions and further guide the depth propagation. Despite the simplicity of the basic idea, our method, PGD, obtains significant improvements on KITTI and nuScenes benchmarks, achieving 1st place out of all monocular vision-only methods while still maintaining real-time efficiency. Code and models will be released at [this https URL](https://github.com/open-mmlab/mmdetection3d).
+
+<div align=center>
+<img src="https://user-images.githubusercontent.com/79644370/143884065-d1a19fdf-bcc0-4249-84cf-b7a85fa1eb2f.png" width="800"/>
+</div>
+
+## Introduction
+
+PGD, also can be regarded as FCOS3D++, is a simple yet effective monocular 3D detector. It enhances the FCOS3D baseline by involving local geometric constraints and improving instance depth estimation.
+
+We release the code and model for both KITTI and nuScenes benchmark, which is a good supplement for the original FCOS3D baseline (only supported on nuScenes).
+
+For clean implementation, our preliminary release supports base models with proposed local geometric constraints and the probabilistic depth representation. We will involve the geometric graph part in the future.
+
+A more extensive study based on FCOS3D and PGD is on-going. Please stay tuned.
+
+## Results and models
+
+### KITTI
+
+|  Backbone   | Lr schd | Mem (GB) | Inf time (fps) | mAP_11 / mAP_40 | Download |
+| :---------: | :-----: | :------: | :------------: | :----: | :------: |
+|[ResNet101](./pgd_r101_caffe_fpn_gn-head_3x4_4x_kitti-mono3d.py)|4x|9.07||18.33 / 13.23|[model](https://download.openmmlab.com/mmdetection3d/v1.0.0_models/pgd/pgd_r101_caffe_fpn_gn-head_3x4_4x_kitti-mono3d/pgd_r101_caffe_fpn_gn-head_3x4_4x_kitti-mono3d_20211022_102608-8a97533b.pth) &#124; [log](https://download.openmmlab.com/mmdetection3d/v1.0.0_models/pgd/pgd_r101_caffe_fpn_gn-head_3x4_4x_kitti-mono3d/pgd_r101_caffe_fpn_gn-head_3x4_4x_kitti-mono3d_20211022_102608.log.json)|
+
+Detailed performance on KITTI 3D detection (3D/BEV) is as follows, evaluated by AP11 and AP40 metric:
+
+|             |     Easy      |    Moderate    |     Hard      |
+|-------------|:-------------:|:--------------:|:-------------:|
+| Car (AP11)  | 24.09 / 30.11 | 18.33 / 23.46  | 16.90 / 19.33 |
+| Car (AP40)  | 19.27 / 26.60 | 13.23 / 18.23  | 10.65 / 15.00 |
+
+Note: mAP represents Car moderate 3D strict AP11 / AP40 results. Because of the limited data for pedestrians and cyclists, the detection performance for these two classes is usually unstable. Therefore, we only list car detection results here. In addition, AP40 is a more recommended metric for reference due to its much better stability.
+
+### NuScenes
+
+|  Backbone   | Lr schd | Mem (GB) | mAP | NDS | Download |
+| :---------: | :-----: | :------: | :----: |:----: | :------: |
+|[ResNet101 w/ DCN](./pgd_r101_caffe_fpn_gn-head_2x16_1x_nus-mono3d.py)|1x|9.20|31.7|39.3|[model](https://download.openmmlab.com/mmdetection3d/v1.0.0_models/pgd/pgd_r101_caffe_fpn_gn-head_2x16_1x_nus-mono3d/pgd_r101_caffe_fpn_gn-head_2x16_1x_nus-mono3d_20211116_195350-f4b5eec2.pth) &#124; [log](https://download.openmmlab.com/mmdetection3d/v1.0.0_models/pgd/pgd_r101_caffe_fpn_gn-head_2x16_1x_nus-mono3d/pgd_r101_caffe_fpn_gn-head_2x16_1x_nus-mono3d_20211116_195350.log.json)|
+|[above w/ finetune](./pgd_r101_caffe_fpn_gn-head_2x16_1x_nus-mono3d_finetune.py)|1x|9.20|34.6|41.1|[model](https://download.openmmlab.com/mmdetection3d/v1.0.0_models/pgd/pgd_r101_caffe_fpn_gn-head_2x16_1x_nus-mono3d_finetune/pgd_r101_caffe_fpn_gn-head_2x16_1x_nus-mono3d_finetune_20211118_093245-fd419681.pth) &#124; [log](https://download.openmmlab.com/mmdetection3d/v1.0.0_models/pgd/pgd_r101_caffe_fpn_gn-head_2x16_1x_nus-mono3d_finetune/pgd_r101_caffe_fpn_gn-head_2x16_1x_nus-mono3d_finetune_20211118_093245.log.json)|
+|above w/ tta|1x|9.20|35.5|41.8||
+|[ResNet101 w/ DCN](./pgd_r101_caffe_fpn_gn-head_2x16_2x_nus-mono3d.py)|2x|9.20|33.6|40.9|[model](https://download.openmmlab.com/mmdetection3d/v1.0.0_models/pgd/pgd_r101_caffe_fpn_gn-head_2x16_2x_nus-mono3d/pgd_r101_caffe_fpn_gn-head_2x16_2x_nus-mono3d_20211112_125314-cb677266.pth) &#124; [log](https://download.openmmlab.com/mmdetection3d/v1.0.0_models/pgd/pgd_r101_caffe_fpn_gn-head_2x16_2x_nus-mono3d/pgd_r101_caffe_fpn_gn-head_2x16_2x_nus-mono3d_20211112_125314.log.json)|
+|[above w/ finetune](./pgd_r101_caffe_fpn_gn-head_2x16_2x_nus-mono3d_finetune.py)|2x|9.20|35.8|42.5|[model](https://download.openmmlab.com/mmdetection3d/v1.0.0_models/pgd/pgd_r101_caffe_fpn_gn-head_2x16_2x_nus-mono3d_finetune/pgd_r101_caffe_fpn_gn-head_2x16_2x_nus-mono3d_finetune_20211114_162135-5ec7c1cd.pth) &#124; [log](https://download.openmmlab.com/mmdetection3d/v1.0.0_models/pgd/pgd_r101_caffe_fpn_gn-head_2x16_2x_nus-mono3d_finetune/pgd_r101_caffe_fpn_gn-head_2x16_2x_nus-mono3d_finetune_20211114_162135.log.json)|
+|above w/ tta|2x|9.20|36.8|43.1||
+
+## Citation
+
+```latex
+@inproceedings{wang2021pgd,
+    title={{Probabilistic and Geometric Depth: Detecting} Objects in Perspective},
+    author={Wang, Tai and Zhu, Xinge and Pang, Jiangmiao and Lin, Dahua},
+    booktitle={Conference on Robot Learning (CoRL) 2021},
+    year={2021}
+}
+# For the baseline version
+@inproceedings{wang2021fcos3d,
+    title={{FCOS3D: Fully} Convolutional One-Stage Monocular 3D Object Detection},
+    author={Wang, Tai and Zhu, Xinge and Pang, Jiangmiao and Lin, Dahua},
+    booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops},
+    year={2021}
+}
+```

configs/pgd/metafile.yml

+Collections:
+  - Name: PGD
+    Metadata:
+      Training Data: KITTI
+      Training Techniques:
+        - SGD
+      Training Resources: 4x TITAN XP
+      Architecture:
+        - PGDHead
+    Paper:
+      URL: https://arxiv.org/abs/2107.14160
+      Title: 'Probabilistic and Geometric Depth: Detecting Objects in Perspective'
+    README: configs/pgd/README.md
+    Code:
+      URL: https://github.com/open-mmlab/mmdetection3d/blob/v1.0.0.dev0/mmdet3d/models/dense_heads/pgd_head.py#17
+      Version: v1.0.0
+
+Models:
+  - Name: pgd_r101_caffe_fpn_gn-head_3x4_4x_kitti-mono3d
+    In Collection: PGD
+    Config: configs/pgd/pgd_r101_caffe_fpn_gn-head_3x4_4x_kitti-mono3d.py
+    Metadata:
+      Training Memory (GB): 9.1
+    Results:
+      - Task: 3D Object Detection
+        Dataset: KITTI
+        Metrics:
+          mAP: 18.33
+    Weights: https://download.openmmlab.com/mmdetection3d/v1.0.0_models/pgd/pgd_r101_caffe_fpn_gn-head_3x4_4x_kitti-mono3d/pgd_r101_caffe_fpn_gn-head_3x4_4x_kitti-mono3d_20211022_102608-8a97533b.pth
+
+  - Name: pgd_r101_caffe_fpn_gn-head_2x16_1x_nus-mono3d
+    In Collection: PGD
+    Config: configs/pgd/pgd_r101_caffe_fpn_gn-head_2x16_1x_nus-mono3d.py
+    Metadata:
+      Training Memory (GB): 9.2
+    Results:
+      - Task: 3D Object Detection
+        Dataset: nuScenes
+        Metrics:
+          mAP: 31.7
+          NDS: 39.3
+    Weights: https://download.openmmlab.com/mmdetection3d/v1.0.0_models/pgd/pgd_r101_caffe_fpn_gn-head_2x16_1x_nus-mono3d/pgd_r101_caffe_fpn_gn-head_2x16_1x_nus-mono3d_20211116_195350-f4b5eec2.pth
+
+  - Name: pgd_r101_caffe_fpn_gn-head_2x16_1x_nus-mono3d_finetune
+    In Collection: PGD
+    Config: configs/pgd/pgd_r101_caffe_fpn_gn-head_2x16_1x_nus-mono3d_finetune.py
+    Metadata:
+      Training Memory (GB): 9.2
+    Results:
+      - Task: 3D Object Detection
+        Dataset: nuScenes
+        Metrics:
+          mAP: 34.6
+          NDS: 41.1
+    Weights: https://download.openmmlab.com/mmdetection3d/v1.0.0_models/pgd/pgd_r101_caffe_fpn_gn-head_2x16_1x_nus-mono3d_finetune/pgd_r101_caffe_fpn_gn-head_2x16_1x_nus-mono3d_finetune_20211118_093245-fd419681.pth
+
+  - Name: pgd_r101_caffe_fpn_gn-head_2x16_2x_nus-mono3d
+    In Collection: PGD
+    Config: configs/pgd/pgd_r101_caffe_fpn_gn-head_2x16_2x_nus-mono3d.py
+    Metadata:
+      Training Memory (GB): 9.2
+    Results:
+      - Task: 3D Object Detection
+        Dataset: nuScenes
+        Metrics:
+          mAP: 33.6
+          NDS: 40.9
+    Weights: https://download.openmmlab.com/mmdetection3d/v1.0.0_models/pgd/pgd_r101_caffe_fpn_gn-head_2x16_2x_nus-mono3d/pgd_r101_caffe_fpn_gn-head_2x16_2x_nus-mono3d_20211112_125314-cb677266.pth
+
+  - Name: pgd_r101_caffe_fpn_gn-head_2x16_2x_nus-mono3d_finetune
+    In Collection: PGD
+    Config: configs/pgd/pgd_r101_caffe_fpn_gn-head_2x16_2x_nus-mono3d_finetune.py
+    Metadata:
+      Training Memory (GB): 9.2
+    Results:
+      - Task: 3D Object Detection
+        Dataset: nuScenes
+        Metrics:
+          mAP: 35.8
+          NDS: 42.5
+    Weights: https://download.openmmlab.com/mmdetection3d/v1.0.0_models/pgd/pgd_r101_caffe_fpn_gn-head_2x16_2x_nus-mono3d_finetune/pgd_r101_caffe_fpn_gn-head_2x16_2x_nus-mono3d_finetune_20211114_162135-5ec7c1cd.pth

configs/pgd/pgd_r101_caffe_fpn_gn-head_2x16_1x_nus-mono3d.py

+_base_ = [
+    '../_base_/datasets/nus-mono3d.py', '../_base_/models/pgd.py',
+    '../_base_/schedules/mmdet_schedule_1x.py', '../_base_/default_runtime.py'
+]
+# model settings
+model = dict(
+    backbone=dict(
+        dcn=dict(type='DCNv2', deform_groups=1, fallback_on_stride=False),
+        stage_with_dcn=(False, False, True, True)),
+    bbox_head=dict(
+        pred_bbox2d=True,
+        group_reg_dims=(2, 1, 3, 1, 2,
+                        4),  # offset, depth, size, rot, velo, bbox2d
+        reg_branch=(
+            (256, ),  # offset
+            (256, ),  # depth
+            (256, ),  # size
+            (256, ),  # rot
+            (),  # velo
+            (256, )  # bbox2d
+        ),
+        loss_depth=dict(type='SmoothL1Loss', beta=1.0 / 9.0, loss_weight=1.0),
+        bbox_coder=dict(
+            type='PGDBBoxCoder',
+            base_depths=((31.99, 21.12), (37.15, 24.63), (39.69, 23.97),
+                         (40.91, 26.34), (34.16, 20.11), (22.35, 13.70),
+                         (24.28, 16.05), (27.26, 15.50), (20.61, 13.68),
+                         (22.74, 15.01)),
+            base_dims=((4.62, 1.73, 1.96), (6.93, 2.83, 2.51),
+                       (12.56, 3.89, 2.94), (11.22, 3.50, 2.95),
+                       (6.68, 3.21, 2.85), (6.68, 3.21, 2.85),
+                       (2.11, 1.46, 0.78), (0.73, 1.77, 0.67),
+                       (0.41, 1.08, 0.41), (0.50, 0.99, 2.52)),
+            code_size=9)),
+    # set weight 1.0 for base 7 dims (offset, depth, size, rot)
+    # 0.05 for 2-dim velocity and 0.2 for 4-dim 2D distance targets
+    train_cfg=dict(code_weight=[
+        1.0, 1.0, 0.2, 1.0, 1.0, 1.0, 1.0, 0.05, 0.05, 0.2, 0.2, 0.2, 0.2
+    ]),
+    test_cfg=dict(nms_pre=1000, nms_thr=0.8, score_thr=0.01, max_per_img=200))
+
+class_names = [
+    'car', 'truck', 'trailer', 'bus', 'construction_vehicle', 'bicycle',
+    'motorcycle', 'pedestrian', 'traffic_cone', 'barrier'
+]
+img_norm_cfg = dict(
+    mean=[103.530, 116.280, 123.675], std=[1.0, 1.0, 1.0], to_rgb=False)
+train_pipeline = [
+    dict(type='LoadImageFromFileMono3D'),
+    dict(
+        type='LoadAnnotations3D',
+        with_bbox=True,
+        with_label=True,
+        with_attr_label=True,
+        with_bbox_3d=True,
+        with_label_3d=True,
+        with_bbox_depth=True),
+    dict(type='Resize', img_scale=(1600, 900), keep_ratio=True),
+    dict(type='RandomFlip3D', flip_ratio_bev_horizontal=0.5),
+    dict(type='Normalize', **img_norm_cfg),
+    dict(type='Pad', size_divisor=32),
+    dict(type='DefaultFormatBundle3D', class_names=class_names),
+    dict(
+        type='Collect3D',
+        keys=[
+            'img', 'gt_bboxes', 'gt_labels', 'attr_labels', 'gt_bboxes_3d',
+            'gt_labels_3d', 'centers2d', 'depths'
+        ]),
+]
+test_pipeline = [
+    dict(type='LoadImageFromFileMono3D'),
+    dict(
+        type='MultiScaleFlipAug',
+        scale_factor=1.0,
+        flip=False,
+        transforms=[
+            dict(type='RandomFlip3D'),
+            dict(type='Normalize', **img_norm_cfg),
+            dict(type='Pad', size_divisor=32),
+            dict(
+                type='DefaultFormatBundle3D',
+                class_names=class_names,
+                with_label=False),
+            dict(type='Collect3D', keys=['img']),
+        ])
+]
+data = dict(
+    samples_per_gpu=2,
+    workers_per_gpu=2,
+    train=dict(pipeline=train_pipeline),
+    val=dict(pipeline=test_pipeline),
+    test=dict(pipeline=test_pipeline))
+# optimizer
+optimizer = dict(
+    lr=0.004, paramwise_cfg=dict(bias_lr_mult=2., bias_decay_mult=0.))
+optimizer_config = dict(
+    _delete_=True, grad_clip=dict(max_norm=35, norm_type=2))
+# learning policy
+lr_config = dict(
+    policy='step',
+    warmup='linear',
+    warmup_iters=500,
+    warmup_ratio=1.0 / 3,
+    step=[8, 11])
+total_epochs = 12
+evaluation = dict(interval=4)
+runner = dict(max_epochs=total_epochs)

configs/pgd/pgd_r101_caffe_fpn_gn-head_2x16_1x_nus-mono3d_finetune.py

+_base_ = './pgd_r101_caffe_fpn_gn-head_2x16_1x_nus-mono3d.py'
+# model settings
+model = dict(
+    train_cfg=dict(code_weight=[
+        1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.05, 0.05, 0.2, 0.2, 0.2, 0.2
+    ]))
+# optimizer
+optimizer = dict(lr=0.002)
+load_from = 'work_dirs/pgd_nus_benchmark_1x/latest.pth'

configs/pgd/pgd_r101_caffe_fpn_gn-head_2x16_2x_nus-mono3d.py

+_base_ = './pgd_r101_caffe_fpn_gn-head_2x16_1x_nus-mono3d.py'
+# learning policy
+lr_config = dict(step=[16, 22])
+total_epochs = 24
+runner = dict(max_epochs=total_epochs)

configs/pgd/pgd_r101_caffe_fpn_gn-head_2x16_2x_nus-mono3d_finetune.py

+_base_ = './pgd_r101_caffe_fpn_gn-head_2x16_2x_nus-mono3d.py'
+# model settings
+model = dict(
+    train_cfg=dict(code_weight=[
+        1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.05, 0.05, 0.2, 0.2, 0.2, 0.2
+    ]))
+# optimizer
+optimizer = dict(lr=0.002)
+load_from = 'work_dirs/pgd_nus_benchmark_2x/latest.pth'

configs/pgd/pgd_r101_caffe_fpn_gn-head_3x4_4x_kitti-mono3d.py

+_base_ = [
+    '../_base_/datasets/kitti-mono3d.py', '../_base_/models/pgd.py',
+    '../_base_/schedules/mmdet_schedule_1x.py', '../_base_/default_runtime.py'
+]
+# model settings
+model = dict(
+    backbone=dict(frozen_stages=0),
+    neck=dict(start_level=0, num_outs=4),
+    bbox_head=dict(
+        num_classes=3,
+        bbox_code_size=7,
+        pred_attrs=False,
+        pred_velo=False,
+        pred_bbox2d=True,
+        use_onlyreg_proj=True,
+        strides=(4, 8, 16, 32),
+        regress_ranges=((-1, 64), (64, 128), (128, 256), (256, 1e8)),
+        group_reg_dims=(2, 1, 3, 1, 16,
+                        4),  # offset, depth, size, rot, kpts, bbox2d
+        reg_branch=(
+            (256, ),  # offset
+            (256, ),  # depth
+            (256, ),  # size
+            (256, ),  # rot
+            (256, ),  # kpts
+            (256, )  # bbox2d
+        ),
+        centerness_branch=(256, ),
+        loss_cls=dict(
+            type='FocalLoss',
+            use_sigmoid=True,
+            gamma=2.0,
+            alpha=0.25,
+            loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0 / 9.0, loss_weight=1.0),
+        loss_dir=dict(
+            type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0),
+        loss_centerness=dict(
+            type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0),
+        use_depth_classifier=True,
+        depth_branch=(256, ),
+        depth_range=(0, 70),
+        depth_unit=10,
+        division='uniform',
+        depth_bins=8,
+        pred_keypoints=True,
+        weight_dim=1,
+        loss_depth=dict(
+            type='UncertainSmoothL1Loss', alpha=1.0, beta=3.0,
+            loss_weight=1.0),
+        bbox_coder=dict(
+            type='PGDBBoxCoder',
+            base_depths=((28.01, 16.32), ),
+            base_dims=((0.8, 1.73, 0.6), (1.76, 1.73, 0.6), (3.9, 1.56, 1.6)),
+            code_size=7)),
+    # set weight 1.0 for base 7 dims (offset, depth, size, rot)
+    # 0.2 for 16-dim keypoint offsets and 1.0 for 4-dim 2D distance targets
+    train_cfg=dict(code_weight=[
+        1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.2, 0.2, 0.2, 0.2, 0.2, 0.2, 0.2,
+        0.2, 0.2, 0.2, 0.2, 0.2, 0.2, 0.2, 0.2, 0.2, 1.0, 1.0, 1.0, 1.0
+    ]),
+    test_cfg=dict(nms_pre=100, nms_thr=0.05, score_thr=0.001, max_per_img=20))
+
+class_names = ['Pedestrian', 'Cyclist', 'Car']
+img_norm_cfg = dict(
+    mean=[103.530, 116.280, 123.675], std=[1.0, 1.0, 1.0], to_rgb=False)
+train_pipeline = [
+    dict(type='LoadImageFromFileMono3D'),
+    dict(
+        type='LoadAnnotations3D',
+        with_bbox=True,
+        with_label=True,
+        with_attr_label=False,
+        with_bbox_3d=True,
+        with_label_3d=True,
+        with_bbox_depth=True),
+    dict(type='Resize', img_scale=(1242, 375), keep_ratio=True),
+    dict(type='RandomFlip3D', flip_ratio_bev_horizontal=0.5),
+    dict(type='Normalize', **img_norm_cfg),
+    dict(type='Pad', size_divisor=32),
+    dict(type='DefaultFormatBundle3D', class_names=class_names),
+    dict(
+        type='Collect3D',
+        keys=[
+            'img', 'gt_bboxes', 'gt_labels', 'gt_bboxes_3d', 'gt_labels_3d',
+            'centers2d', 'depths'
+        ]),
+]
+test_pipeline = [
+    dict(type='LoadImageFromFileMono3D'),
+    dict(
+        type='MultiScaleFlipAug',
+        scale_factor=1.0,
+        flip=False,
+        transforms=[
+            dict(type='RandomFlip3D'),
+            dict(type='Normalize', **img_norm_cfg),
+            dict(type='Pad', size_divisor=32),
+            dict(
+                type='DefaultFormatBundle3D',
+                class_names=class_names,
+                with_label=False),
+            dict(type='Collect3D', keys=['img']),
+        ])
+]
+data = dict(
+    samples_per_gpu=3,
+    workers_per_gpu=3,
+    train=dict(pipeline=train_pipeline),
+    val=dict(pipeline=test_pipeline),
+    test=dict(pipeline=test_pipeline))
+# optimizer
+optimizer = dict(
+    lr=0.001, paramwise_cfg=dict(bias_lr_mult=2., bias_decay_mult=0.))
+optimizer_config = dict(
+    _delete_=True, grad_clip=dict(max_norm=35, norm_type=2))
+# learning policy
+lr_config = dict(
+    policy='step',
+    warmup='linear',
+    warmup_iters=500,
+    warmup_ratio=1.0 / 3,
+    step=[32, 44])
+total_epochs = 48
+runner = dict(type='EpochBasedRunner', max_epochs=48)
+evaluation = dict(interval=2)
+checkpoint_config = dict(interval=8)