[Fix] Update README format

configs/3dssd/README.md

@@ -21,7 +21,15 @@ Currently, there have been many kinds of voxel-based 3D single stage detectors,
 
 We implement 3DSSD and provide the results and checkpoints on KITTI datasets.
 
-```
+Some settings in our implementation are different from the [official implementation](https://github.com/Jia-Research-Lab/3DSSD), which bring marginal differences to the performance on KITTI datasets in our experiments. To simplify and unify the models of our implementation, we skip them in our models. These differences are listed as below:
+1. We keep the scenes without any object while the official code skips these scenes in training. In the official implementation, only 3229 and 3394 samples are used as training and validation sets, respectively. In our implementation, we keep using 3712 and 3769 samples as training and validation sets, respectively, as those used for all the other models in our implementation on KITTI datasets.
+2. We do not modify the decay of `batch normalization` during training.
+3. While using [`DataBaseSampler`](https://github.com/open-mmlab/mmdetection3d/blob/master/mmdet3d/datasets/pipelines/dbsampler.py#L80) for data augmentation, the official code uses road planes as reference to place the sampled objects while we do not.
+4. We perform detection using LIDAR coordinates while the official code uses camera coordinates.
+
+## Citation
+
+```latex
 @inproceedings{yang20203dssd,
     author = {Zetong Yang and Yanan Sun and Shu Liu and Jiaya Jia},
     title = {3DSSD: Point-based 3D Single Stage Object Detector},
@@ -30,15 +38,7 @@ We implement 3DSSD and provide the results and checkpoints on KITTI datasets.
 }
 ```
 
-### Experiment details on KITTI datasets
-
-Some settings in our implementation are different from the [official implementation](https://github.com/Jia-Research-Lab/3DSSD), which bring marginal differences to the performance on KITTI datasets in our experiments. To simplify and unify the models of our implementation, we skip them in our models. These differences are listed as below:
-1. We keep the scenes without any object while the official code skips these scenes in training. In the official implementation, only 3229 and 3394 samples are used as training and validation sets, respectively. In our implementation, we keep using 3712 and 3769 samples as training and validation sets, respectively, as those used for all the other models in our implementation on KITTI datasets.
-2. We do not modify the decay of `batch normalization` during training.
-3. While using [`DataBaseSampler`](https://github.com/open-mmlab/mmdetection3d/blob/master/mmdet3d/datasets/pipelines/dbsampler.py#L80) for data augmentation, the official code uses road planes as reference to place the sampled objects while we do not.
-4. We perform detection using LIDAR coordinates while the official code uses camera coordinates.
-
-## Results
+## Results and models
 
 ### KITTI
 

configs/centerpoint/README.md

@@ -42,7 +42,9 @@ We follow the below style to name config files. Contributors are advised to foll
 
 `{dataset}`: dataset like nus-3d, kitti-3d, lyft-3d, scannet-3d, sunrgbd-3d. We also indicate the number of classes we are using if there exist multiple settings, e.g., kitti-3d-3class and kitti-3d-car means training on KITTI dataset with 3 classes and single class, respectively.
 
-```
+## Citation
+
+```latex
 @article{yin2021center,
   title={Center-based 3D Object Detection and Tracking},
   author={Yin, Tianwei and Zhou, Xingyi and Kr{\"a}henb{\"u}hl, Philipp},
@@ -118,7 +120,7 @@ data = dict(
 
 ```
 
-## Results
+## Results and models
 
 ### CenterPoint
 

configs/dgcnn/README.md

@@ -21,7 +21,9 @@ Point clouds provide a flexible geometric representation suitable for countless
 
 We implement DGCNN and provide the results and checkpoints on S3DIS dataset.
 
-```
+## Citation
+
+```latex
 @article{dgcnn,
   title={Dynamic Graph CNN for Learning on Point Clouds},
   author={Wang, Yue and Sun, Yongbin and Liu, Ziwei and Sarma, Sanjay E. and Bronstein, Michael M. and Solomon, Justin M.},
@@ -32,7 +34,7 @@ We implement DGCNN and provide the results and checkpoints on S3DIS dataset.
 
 **Notice**: We follow the implementations in the original DGCNN paper and a PyTorch implementation of DGCNN [code](https://github.com/AnTao97/dgcnn.pytorch).
 
-## Results
+## Results and models
 
 ### S3DIS
 

configs/dynamic_voxelization/README.md

@@ -21,7 +21,9 @@ Recent work on 3D object detection advocates point cloud voxelization in birds-e
 
 We implement Dynamic Voxelization proposed in  and provide its results and models on KITTI dataset.
 
-```
+## Citation
+
+```latex
 @article{zhou2019endtoend,
     title={End-to-End Multi-View Fusion for 3D Object Detection in LiDAR Point Clouds},
     author={Yin Zhou and Pei Sun and Yu Zhang and Dragomir Anguelov and Jiyang Gao and Tom Ouyang and James Guo and Jiquan Ngiam and Vijay Vasudevan},
@@ -33,7 +35,7 @@ We implement Dynamic Voxelization proposed in  and provide its results and model
 
 ```
 
-## Results
+## Results and models
 
 ### KITTI
 

configs/fcos3d/README.md

@@ -24,7 +24,11 @@ It serves as a baseline built on top of mmdetection and mmdetection3d for 3D det
 
 Currently we first support the benchmark on the large-scale nuScenes dataset, which achieved 1st place out of all the vision-only methods in the [nuScenes 3D detecton challenge](https://www.nuscenes.org/object-detection?externalData=all&mapData=all&modalities=Camera) of NeurIPS 2020.
 
-```
+![demo image](../../resources/browse_dataset_mono.png)
+
+## Citation
+
+```latex
 @inproceedings{wang2021fcos3d,
 	title={{FCOS3D: Fully} Convolutional One-Stage Monocular 3D Object Detection},
 	author={Wang, Tai and Zhu, Xinge and Pang, Jiangmiao and Lin, Dahua},
@@ -40,8 +44,6 @@ Currently we first support the benchmark on the large-scale nuScenes dataset, wh
 }
 ```
 
-![demo image](../../resources/browse_dataset_mono.png)
-
 ## Usage
 
 ### Data Preparation
@@ -67,7 +69,7 @@ Due to the scale and measurements of depth is different from those of other regr
 
 We also provide visualization functions to show the monocular 3D detection results. Simply follow the [documentation](https://mmdetection3d.readthedocs.io/en/latest/1_exist_data_model.html#test-existing-models-on-standard-datasets) and use the `single-gpu testing` command. You only need to add the `--show` flag and specify `--show-dir` to store the visualization results.
 
-## Results
+## Results and models
 
 ### NuScenes
 

configs/free_anchor/README.md

@@ -24,7 +24,10 @@ We implement FreeAnchor in 3D detection systems and provide their first results
 With the implemented `FreeAnchor3DHead`, a PointPillar detector with a big backbone (e.g., RegNet-3.2GF) achieves top performance
 on the nuScenes benchmark.
 
-```
+
+## Citation
+
+```latex
 @inproceedings{zhang2019freeanchor,
   title   =  {{FreeAnchor}: Learning to Match Anchors for Visual Object Detection},
   author  =  {Zhang, Xiaosong and Wan, Fang and Liu, Chang and Ji, Rongrong and Ye, Qixiang},
@@ -91,7 +94,7 @@ model = dict(
         pts=dict(code_weight=[1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.25, 0.25])))
 ```
 
-## Results
+## Results and models
 
 ### PointPillars
 

configs/groupfree3d/README.md

@@ -21,7 +21,9 @@ Recently, directly detecting 3D objects from 3D point clouds has received increa
 
 We implement Group-Free-3D and provide the result and checkpoints on ScanNet datasets.
 
-```
+## Citation
+
+```latex
 @article{liu2021,
   title={Group-Free 3D Object Detection via Transformers},
   author={Liu, Ze and Zhang, Zheng and Cao, Yue and Hu, Han and Tong, Xin},
@@ -30,7 +32,7 @@ We implement Group-Free-3D and provide the result and checkpoints on ScanNet dat
 }
 ```
 
-## Results
+## Results and models
 
 ### ScanNet
 

configs/h3dnet/README.md

@@ -21,7 +21,9 @@ We introduce H3DNet, which takes a colorless 3D point cloud as input and outputs
 
 We implement H3DNet and provide the result and checkpoints on ScanNet datasets.
 
-```
+## 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},
@@ -30,7 +32,7 @@ We implement H3DNet and provide the result and checkpoints on ScanNet datasets.
 }
 ```
 
-## Results
+## Results and models
 
 ### ScanNet
 

configs/imvotenet/README.md

@@ -21,7 +21,9 @@
 
 We implement ImVoteNet and provide the result and checkpoints on SUNRGBD.
 
-```
+## 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},
@@ -31,7 +33,7 @@ We implement ImVoteNet and provide the result and checkpoints on SUNRGBD.
 }
 ```
 
-## Results
+## Results and models
 
 ### SUNRGBD-2D (Stage 1, image branch pre-train)
 

configs/imvoxelnet/README.md

@@ -23,7 +23,9 @@ We implement a monocular 3D detector ImVoxelNet and provide its results and chec
 Results for SUN RGB-D, ScanNet and nuScenes are currently available in ImVoxelNet authors
 [repo](https://github.com/saic-vul/imvoxelnet) (based on mmdetection3d).
 
-```
+## 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},
@@ -32,7 +34,7 @@ Results for SUN RGB-D, ScanNet and nuScenes are currently available in ImVoxelNe
 }
 ```
 
-## Results
+## Results and models
 
 ### KITTI
 

configs/mvxnet/README.md

@@ -19,7 +19,9 @@ Many recent works on 3D object detection have focused on designing neural networ
 
 We implement MVX-Net and provide its results and models on KITTI dataset.
 
-```
+## Citation
+
+```latex
 @inproceedings{sindagi2019mvx,
   title={MVX-Net: Multimodal voxelnet for 3D object detection},
   author={Sindagi, Vishwanath A and Zhou, Yin and Tuzel, Oncel},
@@ -31,7 +33,7 @@ We implement MVX-Net and provide its results and models on KITTI dataset.
 
 ```
 
-## Results
+## Results and models
 
 ### KITTI
 

configs/nuimages/README.md

@@ -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
 

configs/paconv/README.md

@@ -22,7 +22,11 @@ Furthermore, different from the existing point convolution operators whose netwo
 
 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.
+
+## 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},
@@ -32,9 +36,7 @@ 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
 

configs/parta2/README.md

@@ -21,7 +21,9 @@
 
 We implement Part-A^2 and provide its results and checkpoints on KITTI dataset.
 
-```
+## 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},
@@ -31,7 +33,7 @@ We implement Part-A^2 and provide its results and checkpoints on KITTI dataset.
 }
 ```
 
-## Results
+## Results and models
 
 ### KITTI
 

configs/pgd/README.md

@@ -27,7 +27,9 @@ For clean implementation, our preliminary release supports base models with prop
 
 A more extensive study based on FCOS3D and PGD is on-going. Please stay tuned.
 
-```
+## 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},
@@ -43,7 +45,7 @@ A more extensive study based on FCOS3D and PGD is on-going. Please stay tuned.
 }
 ```
 
-## Results
+## Results and models
 
 ### KITTI
 

configs/point_rcnn/README.md

@@ -21,7 +21,9 @@ In this paper, we propose PointRCNN for 3D object detection from raw point cloud
 
 We implement PointRCNN and provide the result with checkpoints on KITTI dataset.
 
-```
+## Citation
+
+```latex
 @inproceedings{Shi_2019_CVPR,
     title = {PointRCNN: 3D Object Proposal Generation and Detection from Point Cloud},
     author = {Shi, Shaoshuai and Wang, Xiaogang and Li, Hongsheng},
@@ -31,7 +33,7 @@ We implement PointRCNN and provide the result with checkpoints on KITTI dataset.
 }
 ```
 
-## Results
+## Results and models
 
 ### KITTI
 

configs/pointnet2/README.md

@@ -21,7 +21,9 @@ Few prior works study deep learning on point sets. PointNet by Qi et al. is a pi
 
 We implement PointNet++ and provide the result and checkpoints on ScanNet and S3DIS datasets.
 
-```
+## Citation
+
+```latex
 @inproceedings{qi2017pointnet++,
   title={PointNet++ deep hierarchical feature learning on point sets in a metric space},
   author={Qi, Charles R and Yi, Li and Su, Hao and Guibas, Leonidas J},
@@ -33,7 +35,7 @@ We implement PointNet++ and provide the result and checkpoints on ScanNet and S3
 
 **Notice**: The original PointNet++ paper used step learning rate schedule. We discovered that cosine schedule achieves much better results and adopt it in our implementations. We also use a larger `weight_decay` factor because we find it consistently improves the performance.
 
-## Results
+## Results and models
 
 ### ScanNet
 

configs/pointpillars/README.md

@@ -21,18 +21,20 @@ Object detection in point clouds is an important aspect of many robotics applica
 
 We implement PointPillars and provide the results and checkpoints on KITTI, nuScenes, Lyft and Waymo datasets.
 
-```
+
+## Citation
+
+```latex
 @inproceedings{lang2019pointpillars,
   title={Pointpillars: Fast encoders for object detection from point clouds},
   author={Lang, Alex H and Vora, Sourabh and Caesar, Holger and Zhou, Lubing and Yang, Jiong and Beijbom, Oscar},
   booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
   pages={12697--12705},
   year={2019}
 }
-
 ```
 
-## Results
+## Results and models
 
 ### KITTI