@@ -104,9 +104,15 @@ Like [MMDetection](https://github.com/open-mmlab/mmdetection) and [MMCV](https:/
### Highlight
**We have renamed the branch `1.1` to `main` and switched the default branch from `master` to `main`. We encourage users to migrate to the latest version, though it comes with some cost. Please refer to [Migration Guide](docs/en/migration.md) for more details.**
In version 1.4, MMDetecion3D refactors the Waymo dataset and accelerates the preprocessing, training/testing setup, and evaluation of Waymo dataset. We also extends the support for camera-based, such as Monocular and BEV, 3D object detection models on Waymo. A detailed description of the Waymo data information is provided [here](https://mmdetection3d.readthedocs.io/en/latest/advanced_guides/datasets/waymo.html).
We have constructed a comprehensive LiDAR semantic segmentation benchmark on SemanticKITTI, including Cylinder3D, MinkUNet and SPVCNN methods. Noteworthy, the improved MinkUNetv2 can achieve 70.3 mIoU on the validation set of SemanticKITTI. We have also supported the training of BEVFusion and an occupancy prediction method, TPVFomrer, in our `projects`. More new features about 3D perception are on the way. Please stay tuned!
Besides, in version 1.4, MMDetection3D provides [Waymo-mini](https://download.openmmlab.com/mmdetection3d/data/waymo_mmdet3d_after_1x4/waymo_mini.tar.gz) to help community users get started with Waymo and use it for quick iterative development.
**v1.4.0** was released in 8/1/2024：
- Support the training of [DSVT](<(https://arxiv.org/abs/2301.06051)>) in `projects`
- Support [Nerf-Det](https://arxiv.org/abs/2307.14620) in `projects`
In this technical report, we present our solution, dubbed MV-FCOS3D++, for the Camera-Only 3D Detection track in Waymo Open Dataset Challenge 2022. For multi-view camera-only 3D detection, methods based on bird-eye-view or 3D geometric representations can leverage the stereo cues from overlapped regions between adjacent views and directly perform 3D detection without hand-crafted post-processing. However, it lacks direct semantic supervision for 2D backbones, which can be complemented by pretraining simple monocular-based detectors. Our solution is a multi-view framework for 4D detection following this paradigm. It is built upon a simple monocular detector FCOS3D++, pretrained only with object annotations of Waymo, and converts multi-view features to a 3D grid space to detect 3D objects thereon. A dual-path neck for single-frame understanding and temporal stereo matching is devised to incorporate multi-frame information. Our method finally achieves 49.75% mAPL with a single model and wins 2nd place in the WOD challenge, without any LiDAR-based depth supervision during training. The code will be released at [this https URL](https://github.com/Tai-Wang/Depth-from-Motion).
Regrettably, we are unable to provide the pre-trained model weights due to [Waymo Dataset License Agreement](https://waymo.com/open/terms/), so we only provide the training logs as shown above.
## Citation
```latex
@article{wang2022mvfcos3d++,
title={{MV-FCOS3D++: Multi-View} Camera-Only 4D Object Detection with Pretrained Monocular Backbones},
author={Wang, Tai and Lian, Qing and Zhu, Chenming and Zhu, Xinge and Zhang, Wenwei},
Regrettably, we are unable to provide the pre-trained model weights due to [Waymo Dataset License Agreement](https://waymo.com/open/terms/), so we only provide the training logs as shown above.
