releasing SPT tutorial slides and notebook

Author: drprojects

README.md

@@ -121,6 +121,9 @@ Due to its lightweight backbone and scalable formulation, SuperCluster can proce
 
 ## 📰  Updates
 
+- **27.06.2024** Released 🧑‍🏫 tutorial 
+[slides](media/superpoint_transformer_tutorial.pdf) and 
+[notebook](notebooks/superpoint_transformer_tutorial.ipynb). 
 - **21.06.2024** [Damien](https://github.com/drprojects) will be giving a 
 **🧑‍🏫 tutorial on Superpoint Transformer on 📅 27.06.2024 at 1pm CEST**. 
 Make sure to come if you want to gain some hands-on experience with the project !
@@ -224,9 +227,9 @@ named `spt`.
 
 ```
 
-> **Note**: See the [Datasets page](docs/datasets.md) for further details on `data/``. 
+> **Note**: See the [Datasets page](docs/datasets.md) for further details on `data/`. 
 
-> **Note**: See the [Logs page](docs/logging.md) for further details on `logs/``. 
+> **Note**: See the [Logs page](docs/logging.md) for further details on `logs/`. 
 
 <br>
 

docs/datasets.md

@@ -23,6 +23,12 @@ for more details.
 <summary><b>S3DIS data directory structure.</b></summary>
 <br><br>
 
+<p align="center">
+    <img width="32%" src="../media/s3dis/input.png">
+    <img width="32%" src="../media/s3dis/sem_gt.png">
+    <img width="32%" src="../media/s3dis/pano_gt.png">
+</p>
+
 ```
 └── data
     └── s3dis                                                     # Structure for S3DIS
@@ -51,6 +57,12 @@ for more details.
 <summary><b>ScanNetV2 data directory structure.</b></summary>
 <br><br>
 
+<p align="center">
+    <img width="32%" src="../media/scannet/input.png">
+    <img width="32%" src="../media/scannet/sem_gt.png">
+    <img width="32%" src="../media/scannet/pano_gt.png">
+</p>
+
 ```
 └── data
     └─── scannet                                     # Structure for ScanNetV2
@@ -79,6 +91,12 @@ for more details.
 <summary><b>KITTI-360 data directory structure.</b></summary>
 <br><br>
 
+<p align="center">
+    <img width="32%" src="../media/kitti360/input.png">
+    <img width="32%" src="../media/kitti360/sem_gt.png">
+    <img width="32%" src="../media/kitti360/pano_gt.png">
+</p>
+
 ```
 └── data
     └─── kitti360                                     # Structure for KITTI-360
@@ -103,6 +121,12 @@ for more details.
 <summary><b>DALES data directory structure.</b></summary>
 <br><br>
 
+<p align="center">
+    <img width="32%" src="../media/dales/input.png">
+    <img width="32%" src="../media/dales/sem_gt.png">
+    <img width="32%" src="../media/dales/pano_gt.png">
+</p>
+
 ```
 └── data
     └── dales                                         # Structure for DALES
@@ -228,11 +252,11 @@ the `num_classes` you define in your `YourDataset`.
 - **All labels $[0, C - 1]$ are assumed to be present in your dataset**. As 
 such, they will all be used in metrics and losses computation.
 - A point with the **$C$ label will be considered void/ignored/unlabeled** 
-(whichever you call it). As such, it will be excluded from from metrics and 
+(whichever you call it). As such, it will be excluded from metrics and 
 losses computation
 
 Hence, make sure the **output of your `YourDataset.read_single_raw_cloud()` 
-reader method never returns labels outside of your $[0, C]$ range**. Besides, 
+reader method never returns labels outside your $[0, C]$ range**. Besides, 
 if some labels in $[0, C - 1]$ are not useful to you (ie absent from your 
 dataset), we recommend you remap your labels to a new $[0, C' - 1]$ range
 (`torch_geometric.nn.pool.consecutive.consecutive_cluster` can help you with 

docs/visualization.md

@@ -20,10 +20,6 @@
     "file_path = os.path.dirname(os.path.abspath(''))  # for .ipynb notebook\n",
     "sys.path.append(file_path)\n",
     "\n",
-    "# Necessary for advanced config parsing with hydra and omegaconf\n",
-    "from omegaconf import OmegaConf\n",
-    "OmegaConf.register_new_resolver(\"eval\", eval)\n",
-    "\n",
     "import hydra\n",
     "from src.utils import init_config\n",
     "import torch\n",
@@ -138,12 +134,7 @@
     "\n",
     "# Load pretrained weights from a checkpoint file\n",
     "if ckpt_widget.value is not None:\n",
-    "    if task_widget.value == 'semantic':\n",
-    "        model = model.__class__.load_from_checkpoint(cfg.ckpt_path, net=model.net, criterion=model.criterion)    \n",
-    "    elif task_widget.value == 'panoptic':\n",
-    "        model = model.__class__.load_from_checkpoint(cfg.ckpt_path, net=model.net, edge_affinity_head=model.edge_affinity_head, partitioner=model.partitioner, criterion=model.criterion)\n",
-    "    else:\n",
-    "        raise ValueError(f\"Unknown task '{task_widget.value}'\")\n",
+    "    model = model._load_from_checkpoint(cfg.ckpt_path)\n",
     "\n",
     "# Move model to selected device\n",
     "model = model.eval().to(device_widget.value)"
@@ -475,9 +466,9 @@
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": "Python [conda env:spt_test] *",
+   "display_name": "Python [conda env:spt] *",
    "language": "python",
-   "name": "conda-env-spt_test-py"
+   "name": "conda-env-spt-py"
   },
   "language_info": {
    "codemirror_mode": {
@@ -489,7 +480,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.8.18"
+   "version": "3.8.19"
   }
  },
  "nbformat": 4,