|
|
|
|
|
TITLE = """<h1 align="center" id="space-title">GridNet-HD leaderboard</h1>""" |
|
|
|
|
|
|
|
|
INTRODUCTION_TEXT = """ |
|
|
This public Hugging Face Leaderboard evaluates the effectiveness of LiDAR–image fusion methods on the [GridNet-HD dataset](https://huggingface.co/datasets/heig-vd-geo/GridNet-HD) for 3D semantic segmentation of power line infrastructure. |
|
|
The dataset is associated with the following paper: |
|
|
|
|
|
> **Title**: GridNet-HD: A High-Resolution Multi-Modal Dataset for LiDAR-Image Fusion on Power Line Infrastructure |
|
|
|
|
|
> **Authors**: Masked for instance |
|
|
|
|
|
> **Conference**: Submitted to NeurIPS 2025 |
|
|
|
|
|
""" |
|
|
|
|
|
|
|
|
BENCHMARKS_TEXT = """ |
|
|
## How it works |
|
|
|
|
|
Please respect the files structure from the [GridNet-HD repository](https://huggingface.co/datasets/heig-vd-geo/GridNet-HD) |
|
|
|
|
|
``` |
|
|
dataset-root/ |
|
|
├── t1z5b/ |
|
|
│ ├── images/ # RGB images (.JPG) |
|
|
│ ├── masks/ # Semantic segmentation masks (.png, single-channel label) |
|
|
│ ├── lidar/ # LiDAR point cloud (.las format with field "ground_truth") |
|
|
│ └── pose/ # Camera poses and intrinsics (text files) |
|
|
├── t1z6a/ |
|
|
│ ├── ... |
|
|
├── ... |
|
|
├── split.json # JSON file specifying the train/test split |
|
|
└── README.md |
|
|
``` |
|
|
|
|
|
Available test areas are listed in `split.json` : t1z4, t1z5a, t1z7, t3z1, t3z2, t3z5, t5a2, t6z1, t6z5 |
|
|
|
|
|
For example, you can run SPT baseline as described in [SPT_GridNet-HD_baseline#usage-examples](https://huggingface.co/heig-vd-geo/SPT_GridNet-HD_baseline#usage-examples) : |
|
|
|
|
|
```python |
|
|
python inference.py --mode inference --split test --weights path/to/model.ckpt --root_dir /path/to/data/gridnet/raw |
|
|
``` |
|
|
|
|
|
Then, resulting las file with your added classification field must be converted to NPZ by providing only the classification term, like : |
|
|
|
|
|
```python |
|
|
def create_npz(field_name:str="classification", root_dir:str="GridNet-HD"): |
|
|
|
|
|
# Iterate over each area directory |
|
|
for area_name in ["t1z4", "t1z5a", "t1z7", "t3z1", "t3z2", "t3z5", "t5a2", "t6z1", "t6z5"]: |
|
|
area_path = os.path.join(root_dir, area_name) |
|
|
lidar_path = os.path.join(area_path, "lidar") |
|
|
|
|
|
# Check if the lidar directory exists |
|
|
if not os.path.isdir(lidar_path): |
|
|
continue |
|
|
|
|
|
# Find the LAS file in the lidar directory |
|
|
las_files = [f for f in os.listdir(lidar_path) if f.lower().endswith(".las")] |
|
|
if not las_files: |
|
|
print(f"No LAS file found in {lidar_path}") |
|
|
continue |
|
|
|
|
|
las_file_path = os.path.join(lidar_path, las_files[0]) |
|
|
print(las_file_path) |
|
|
# Read the LAS file |
|
|
try: |
|
|
las = laspy.read(las_file_path) |
|
|
except Exception as e: |
|
|
print(f"Error reading {las_file_path}: {e}") |
|
|
continue |
|
|
|
|
|
if field_name not in las.point_format.dimension_names: |
|
|
print(f"Error: field '{field_name}' not found in LAS file!") |
|
|
continue |
|
|
|
|
|
# Extract the ground_truth data |
|
|
classif_data = las[field_name].astype(np.uint8) |
|
|
las = None |
|
|
|
|
|
# Save the ground_truth data as a compressed .npz file |
|
|
npz_filename = f"{area_name}.npz" |
|
|
npz_path = os.path.join(root_dir, "npz", field_name, npz_filename) |
|
|
# ensure dir exist |
|
|
os.makedirs(os.path.dirname(npz_path), exist_ok=True) |
|
|
np.savez_compressed(npz_path, data=classif_data.astype(np.uint8)) |
|
|
``` |
|
|
|
|
|
Resulting NPZ files can be uploaded to the leaderboard using our `Submit Eval` form. |
|
|
|
|
|
|
|
|
## How to reproduce our results |
|
|
|
|
|
Please follow instructions on dedicated git repository for models running on this dataset: |
|
|
- Baseline based on image segmentation and reprojection into LiDAR: [ImageVote baseline](https://huggingface.co/heig-vd-geo/ImageVote_GridNet-HD_baseline) |
|
|
- Baseline based on LiDAR 3D segmentation directly using Superpoint Trasnformer (SPT): [SPT baseline](https://huggingface.co/heig-vd-geo/SPT_GridNet-HD_baseline) |
|
|
- Baseline based on late fusion between softmax logits from SPT and ImageVote: [LateFusionMLP baseline](https://huggingface.co/heig-vd-geo/LateFusionMLP_GridNet-HD_baseline) |
|
|
|
|
|
|
|
|
""" |
|
|
|
|
|
EVALUATION_QUEUE_TEXT = """ |
|
|
## Some good practices before submitting your results to the leaderboard. |
|
|
|
|
|
Make sure you convert your LAS files to NPZ before submitting them to the leaderboard. You can use the function `create_npz` in the About section to do this. |
|
|
|
|
|
You can upload one or several NPZ files for each area of the Test dataset. The leaderboard will compare your results with the ground truth data for each area. |
|
|
You must keep points in the same order inside each NPZ file. |
|
|
""" |
|
|
|
|
|
CITATION_BUTTON_LABEL = "Copy the following snippet to cite these results" |
|
|
CITATION_BUTTON_TEXT = r""" |
|
|
GridNet-HD: A High-Resolution Multi-Modal Dataset for LiDAR-Image Fusion on Power Line Infrastructure |
|
|
Masked Authors |
|
|
Submitted to NeurIPS 2025. |
|
|
""" |
|
|
|
