Update README.md

README.md

@@ -1,3 +1,34 @@
----
-license: mit
----
+---
+license: mit
+datasets:
+- labelmaker/arkit_labelmaker
+base_model:
+- Pointcept/PointTransformerV3
+---
+
+# Models Trained with [ARKit LabelMaker](https://huggingface.co/datasets/labelmaker/arkit_labelmaker)
+
+1. Please check out out project page [labelmaker.org](https://labelmaker.org/). 😁
+2. The major dataset is [here](https://huggingface.co/datasets/labelmaker/arkit_labelmaker) and the processing code and preprocessed dataset for PTv3 is [here](https://huggingface.co/datasets/labelmaker/ALC_pointcept).
+3. Training log tensorboard [link](https://huggingface.co/labelmaker/PTv3-ARKit-LabelMaker/tensorboard)
+4. We made some modification on [Pointcept](https://github.com/quantaji/LabelMaker-Pointcept). You can also download per-experiment code base in each experiment folder.
+
+## 1. 3D Semantic Segmentation
+
+Model: PointTransformer V3 + PPT.
+Additional dataset: ScanNet++, Structured3D.
+
+| Task                   | ScanNet     | ScanNet200  | ScanNet mIoU | ScanNet200 mIoU | ckpt                                                                                   | config                                                                      | folder                                                                                                                            |
+| ---------------------- | ----------- | ----------- | ------------ | --------------- | -------------------------------------------------------------------------------------- | --------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------- |
+| Eval ScanNet on val    | train       | train + val | 79.1         | -               | [link](./scannet/semseg-pt-v3m1-1-ppt-extreme-with-alc/model/model_last.pth)           | [link](./scannet/semseg-pt-v3m1-1-ppt-extreme-with-alc/config.py)           | [link](https://huggingface.co/labelmaker/PTv3-ARKit-LabelMaker/tree/main/scannet/semseg-pt-v3m1-1-ppt-extreme-with-alc)           |
+| Eval ScanNet200 on val | train + val | train       | -            | 37.5            | [link](./scannet200/semseg-pt-v3m1-1-ppt-extreme-with-alc/model/model_last.pth)        | [link](./scannet200/semseg-pt-v3m1-1-ppt-extreme-with-alc/config.py)        | [link](https://huggingface.co/labelmaker/PTv3-ARKit-LabelMaker/tree/main/scannet200/semseg-pt-v3m1-1-ppt-extreme-with-alc)        |
+| Eval on both on test   | train + val | train + val | 79.8         | 41.4            | [link](./scannet200/semseg-pt-v3m1-1-ppt-extreme-with-alc-submit/model/model_last.pth) | [link](./scannet200/semseg-pt-v3m1-1-ppt-extreme-with-alc-submit/config.py) | [link](https://huggingface.co/labelmaker/PTv3-ARKit-LabelMaker/tree/main/scannet200/semseg-pt-v3m1-1-ppt-extreme-with-alc-submit) |
+
+## 2. 3D Instance Segmentation
+Pre-trained model checkpoints: PointTransformer V3 + PPT.
+Training dataset: ScanNet/ScanNet200 training split.
+
+| Task       | val mAP | ckpt                                                                            | config                                                               | folder                                                                                                                     |
+| ---------- | ------- | ------------------------------------------------------------------------------- | -------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------- |
+| ScanNet    | 41.4    | [link](./scannet/insseg-pointgroup-v1m1-pt-v3m1-ppt-ft/model/model_last.pth)    | [link](./scannet/insseg-pointgroup-v1m1-pt-v3m1-ppt-ft/config.py)    | [link](https://huggingface.co/labelmaker/PTv3-ARKit-LabelMaker/tree/main/scannet/insseg-pointgroup-v1m1-pt-v3m1-ppt-ft/)   |
+| ScanNet200 | 24.5    | [link](./scannet200/insseg-pointgroup-v1m1-pt-v3m1-ppt-ft/model/model_last.pth) | [link](./scannet200/insseg-pointgroup-v1m1-pt-v3m1-ppt-ft/config.py) | [link](https://huggingface.co/labelmaker/PTv3-ARKit-LabelMaker/tree/main/scannet200/insseg-pointgroup-v1m1-pt-v3m1-ppt-ft) |