feat(python): run dufomap in python now!

updates:
* main.py script to run.
* poster for dufomap.
* Dockerfile usage.

Author: Kin-Zhang

.gitignore

@@ -58,3 +58,5 @@ docs/html/*
 # coverage report
 coverage_report/*
 .coverage/
+
+*.pcd

README.md

@@ -3,7 +3,7 @@
 </p>
 
 [![arXiv](https://img.shields.io/badge/arXiv-2403.01449-b31b1b?logo=arxiv&logoColor=white)](https://arxiv.org/abs/2403.01449)
-[![page](https://img.shields.io/badge/Web-Page-green)](https://KTH-RPL.github.io/dufomap) [video coming soon] [poster coming soon]. Accepted by RA-L'24.
+[![page](https://img.shields.io/badge/Project-Page-green)](https://KTH-RPL.github.io/dufomap) [![poster](https://img.shields.io/badge/RAL2024|Poster-6495ed?style=flat&logo=Shotcut&logoColor=wihte)](https://mit-spark.github.io/Longterm-Perception-WS/assets/proceedings/DUFOMap/poster.pdf) [video coming soon]
 
 Quick Demo: Run with the **same parameter setting** without tuning for different sensor (e.g 16, 32, 64, and 128 channel LiDAR and Livox-series mid360), the following shows the data collected from:
 
@@ -12,20 +12,26 @@ Quick Demo: Run with the **same parameter setting** without tuning for different
 | ![](assets/imgs/dufomap_leica.gif) | ![](assets/imgs/doals_train_128.gif) | ![](assets/imgs/two_floor_mid360.gif) |
 <!-- | ------- | ------- | ------- | -->
 
+🚀 2024-11-20: Update dufomap Python API from [SeFlow](https://github.com/KTH-RPL/SeFlow) try it now! `pip install dufomap` and run `python main.py --data_dir data/00` to get the cleaned map directly. Support all >=Python 3.8 in Windows and Linux. Please extract your own data to unified format first follow [this wiki page](https://kth-rpl.github.io/DynamicMap_Benchmark/data/creation/#custom-data).
+
 ## 0. Setup
 
 ```bash
 sudo apt update && sudo apt install gcc-10 g++-10
 sudo apt install libtbb-dev liblz4-dev
 ```
 
-Dockerfile will be soon available.
 
 Clone quickly and init submodules:
 ```bash
 git clone --recursive -b main --single-branch https://github.com/KTH-RPL/dufomap.git
 ```
 
+Or you can directly build docker image through our [Dockerfile](Dockerfile):
+```bash
+docker build -t dufomap .
+```
+
 ## 1. Build & Run
 
 Build:
@@ -34,7 +40,7 @@ Build:
 cmake -B build -D CMAKE_CXX_COMPILER=g++-10 && cmake --build build
 ```
 
-Prepare Data: Teaser data (KITTI 00: 384.4Mb) can be downloaded via follow commands, more data detail can be found in the [dataset section](https://github.com/KTH-RPL/DynamicMap_Benchmark?tab=readme-ov-file#dataset--scripts) or format your own dataset follow [custom dataset section](https://github.com/KTH-RPL/DynamicMap_Benchmark/blob/master/scripts/README.md#custom-dataset).
+Prepare Data: Teaser data (KITTI 00: 384.4Mb) can be downloaded via follow commands, more data detail can be found in the [dataset section](https://kth-rpl.github.io/DynamicMap_Benchmark/data) or format your own dataset follow [custom dataset section](https://kth-rpl.github.io/DynamicMap_Benchmark/data/creation/#custom-data).
 
 ```bash
 wget https://zenodo.org/records/8160051/files/00.zip -p data

main.py

@@ -0,0 +1,74 @@
+"""
+# Created: 2024-11-20 13:11
+# Copyright (C) 2024-now, RPL, KTH Royal Institute of Technology
+# Author: Qingwen Zhang  (https://kin-zhang.github.io/)
+#
+# This file is part of DUFOMap (https://github.com/KTH-RPL/dufomap) and 
+# DynamicMap Benchmark (https://github.com/KTH-RPL/DynamicMap_Benchmark) projects.
+# If you find this repo helpful, please cite the respective publication as 
+# listed on the above website.
+
+# Description: Output Cleaned Map through Python API.
+"""
+from pathlib import Path
+import os, fire, time
+import numpy as np
+from tqdm import tqdm
+
+from dufomap import dufomap
+from dufomap.utils import pcdpy3
+def inv_pose_matrix(pose):
+    inv_pose = np.eye(4)
+    inv_pose[:3, :3] = pose[:3, :3].T
+    inv_pose[:3, 3] = -pose[:3, :3].T.dot(pose[:3, 3])
+    return inv_pose
+class DynamicMapData:
+    def __init__(self, directory):
+        self.scene_id = directory.split("/")[-1]
+        self.directory = Path(directory) / "pcd"
+        self.pcd_files = [os.path.join(self.directory, f) for f in sorted(os.listdir(self.directory)) if f.endswith('.pcd')]
+
+    def __len__(self):
+        return len(self.pcd_files)
+    
+    def __getitem__(self, index_):
+        res_dict = {
+            'scene_id': self.scene_id,
+            'timestamp': self.pcd_files[index_].split("/")[-1].split(".")[0],
+        }
+        pcd_ = pcdpy3.PointCloud.from_path(self.pcd_files[index_])
+        pc0 = pcd_.np_data[:,:3]
+        res_dict['pc'] = pc0.astype(np.float32)
+        res_dict['pose'] = list(pcd_.viewpoint)
+        return res_dict
+
+def main_vis(
+    data_dir: str = "/home/kin/data/00",
+):
+    dataset = DynamicMapData(data_dir)
+
+    # STEP 0: initialize 
+    mydufo = dufomap(0.1, 0.2, 2, num_threads=12) # resolution, d_s, d_p same with paper.
+    cloud_acc = np.zeros((0, 3), dtype=np.float32)
+    for data_id in (pbar := tqdm(range(0, len(dataset)),ncols=100)):
+        data = dataset[data_id]
+        now_scene_id = data['scene_id']
+        pbar.set_description(f"id: {data_id}, scene_id: {now_scene_id}, timestamp: {data['timestamp']}")
+
+        # STEP 1: integrate point cloud into dufomap
+        mydufo.run(data['pc'], data['pose'], cloud_transform = False) # since pc already in world frame
+        cloud_acc = np.concatenate((cloud_acc, data['pc']), axis=0)
+    
+    # STEP 2: propagate
+    mydufo.oncePropagateCluster(if_propagate=True, if_cluster=False)
+    # STEP 3: Map results
+    mydufo.outputMap(cloud_acc, voxel_map=False)
+    # NOTE(Qingwen): You can also save voxeled map directly based on the resolution we set before:
+    # mydufo.outputMap(cloud_acc, voxel_map=True)
+    
+    mydufo.printDetailTiming()
+
+if __name__ == "__main__":
+    start_time = time.time()
+    fire.Fire(main_vis)
+    print(f"Time used: {time.time() - start_time:.2f} s")