docs(README): leaderboard submission process and fix typo.

docs(workflow): github issue stale auto

perf(vis): faster label view by paint

Author: Kin-Zhang

.github/issue_stale.yaml

@@ -0,0 +1,16 @@
+# Number of days of inactivity before an issue becomes stale
+daysUntilStale: 60
+# Number of days of inactivity before a stale issue is closed
+daysUntilClose: 7
+# Issues with these labels will never be considered stale
+exemptLabels:
+  - backlog
+# Label to use when marking an issue as stale
+staleLabel: stale
+# Comment to post when marking an issue as stale. Set to `false` to disable
+markComment: >
+  This issue has been automatically marked as stale because it has not had
+  recent activity. It will be closed if no further activity occurs. Thank you
+  for your contributions.
+# Comment to post when closing a stale issue. Set to `false` to disable
+closeComment: false

README.md

@@ -6,7 +6,8 @@ SeFlow: A Self-Supervised Scene Flow Method in Autonomous Driving
 [poster comming soon]
 [video coming soon]
 
-2024/07/16 17:18: Most of codes already uploaded and tested. You can to try training directly by downloading demo data. The process script will be public when the paper published. 
+2024/07/16 17:18: Most of codes already uploaded and tested. You can to try training directly by [downloading](https://zenodo.org/records/12751363) demo data or pretrained weight for evaluation. 
+The process script will be public when the paper published. 
 
 Pre-trained weights for models are available in [Zenodo](https://zenodo.org/records/12751363) link. Check usage in [2. Evaluation](#2-evaluation) or [3. Visualization](#3-visualization).
 
@@ -40,7 +41,7 @@ You can try following methods in our code without any effort to make your own be
 
 </details>
 
-💡: Want to learn how to add your own network in this structure? Check [Contribute](assets/README.md#contribute) section and know more about the code.
+💡: Want to learn how to add your own network in this structure? Check [Contribute](assets/README.md#contribute) section and know more about the code. Fee free to pull request!
 
 ## 0. Setup
 
@@ -117,12 +118,12 @@ Or you can directly download the pre-trained weight from [Zenodo](https://zenodo
 
 You can also train the supervised baseline model in our paper with the following command. [Runtime: Around 10 hours in 4x A100 GPUs.] 
 ```bash
-python 1_train.py model=fastflow3d lr=2e-4 epochs=20 batch_size=16 loss_fn=deflowLoss
-python 1_train.py model=deflow lr=2e-4 epochs=20 batch_size=16 loss_fn=ff3dLoss
+python 1_train.py model=fastflow3d lr=2e-4 epochs=20 batch_size=16 loss_fn=ff3dLoss
+python 1_train.py model=deflow lr=2e-4 epochs=20 batch_size=16 loss_fn=deflowLoss
 ```
 
 Note: You may found the different settings in the paper that is all methods are enlarge learning rate to 2e-4 and decrease the epochs to 20 for faster converge (Through analysis, we also found it had better performance). 
-However, we kept the setting on lr=2e-6 and 50 epochs in the paper experiment for fair comparison with ZeroFlow where we directly use their provided weights etc.
+However, we kept the setting on lr=2e-6 and 50 epochs in the paper experiment for the fair comparison with ZeroFlow where we directly use their provided weights.
 
 ## 2. Evaluation
 
@@ -142,6 +143,9 @@ python 2_eval.py checkpoint=/home/kin/seflow_best.ckpt av2_mode=test leaderboard
 python 2_eval.py checkpoint=/home/kin/seflow_best.ckpt av2_mode=test leaderboard_version=2
 ```
 
+And the terminal will output the command for you to submit the result to the online leaderboard. You can follow [this section for evalai](https://github.com/KTH-RPL/DeFlow?tab=readme-ov-file#2-evaluation).
+
+Check all detailed result files (presented in our paper Table 1) in [this discussion](https://github.com/KTH-RPL/DeFlow/discussions/2).
 
 ## 3. Visualization
 

assets/README.md

@@ -70,14 +70,14 @@ python -c "from assets.cuda.chamfer3D import nnChamferDis;print('successfully im
    The cuda version: `pytorch::pytorch-cuda` and `nvidia::cudatoolkit` need be same. [Reference link](https://github.com/pytorch/pytorch/issues/90673#issuecomment-1563799299)
 
 
-3. In cluster have error: `pandas ImportError: /lib64/libstdc++.so.6: version `GLIBCXX_3.4.29' not found`
+3. In cluster have error: `pandas ImportError: /lib64/libstdc++.so.6: version 'GLIBCXX_3.4.29' not found`
     Solved by `export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/proj/berzelius-2023-154/users/x_qinzh/mambaforge/lib`
 
 
 ## Contribute
 
 If you want to contribute to new model, here are tips you can follow:
-1. Dataloader: we believe all data could be process to `.h5`, we named as different scene and inside a scene, the key of each data is timestamp.
+1. Dataloader: we believe all data could be process to `.h5`, we named as different scene and inside a scene, the key of each data is timestamp. Check [dataprocess/README.md](../dataprocess/README.md#process) for more details.
 2. Model: All model files can be found [here: scripts/network/models](../scripts/network/models). You can view deflow and fastflow3d to know how to implement a new model.
 3. Loss: All loss files can be found [here: scripts/network/loss_func.py](../scripts/network/loss_func.py). There are three loss functions already inside the file, you can add a new one following the same pattern.
 4. Training: Once you have implemented the model, you can add the model to the config file [here: conf/model](../conf/model) and train the model using the command `python 1_train.py model=your_model_name`. One more note here may: if your res_dict from model output is different, you may need add one pattern in `def training_step` and `def validation_step`.

assets/slurm/dufolabel_sbatch.py

@@ -0,0 +1,58 @@
+"""
+# Created: 2023-11-30 17:02
+# Copyright (C) 2023-now, RPL, KTH Royal Institute of Technology
+# Author: Qingwen Zhang  (https://kin-zhang.github.io/)
+#
+# 
+# Description: Write sbatch files for DUFO jobs on cluster (SLURM), no GPU needed
+# Reference: 
+#   * ZeroFlow data sbatch: https://github.com/kylevedder/zeroflow/blob/master/data_prep_scripts/split_nsfp_jobs_sbatch.py
+
+# Run with following commands (only train need to be processed with dufo label)
+- python assets/slurm/dufolabel_sbatch.py --split 50 --total 700 --interval 1 --data_dir /home/kin/data/av2/preprocess/sensor/train --data_mode train
+- python assets/slurm/dufolabel_sbatch.py --split 100 --total 800  --interval 2 --data_dir /proj/berzelius-2023-154/users/x_qinzh/dataset/waymo/fix_preprocess/train
+"""
+
+import fire, time, os
+def main(
+    data_dir: str = "/proj/berzelius-2023-154/users/x_qinzh/av2/preprocess/sensor/train",
+    split: int = 50,
+    total: int = 2001,
+    interval: int = 1,
+):
+    # +1 因为range是左闭右开
+    for i in range(0, total+1, split):
+        scene_range = [i , min(i + split, total+1)]
+        print(scene_range)
+        sbatch_file_content = \
+        f"""#!/bin/bash
+#SBATCH -J pack_{scene_range[0]}_{scene_range[1]}
+#SBATCH --gpus 0
+#SBATCH --cpus-per-task 32
+#SBATCH --mem 64G
+#SBATCH --mincpus=32
+#SBATCH -t 1-00:00:00
+#SBATCH --mail-type=END,FAIL
+#SBATCH --mail-user=qingwen@kth.se
+#SBATCH --output /proj/berzelius-2023-154/users/x_qinzh/seflow/logs/slurm/0_lidar/%J_{scene_range[0]}_{scene_range[1]}.out
+#SBATCH --error  /proj/berzelius-2023-154/users/x_qinzh/seflow/logs/slurm/0_lidar/%J_{scene_range[0]}_{scene_range[1]}.err
+
+cd /proj/berzelius-2023-154/users/x_qinzh/seflow
+export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/proj/berzelius-2023-154/users/x_qinzh/mambaforge/lib
+
+/proj/berzelius-2023-154/users/x_qinzh/mambaforge/envs/seflow/bin/python 0_process.py \\
+    --data_dir {data_dir} \\
+    --interval {interval} \\
+    --scene_range {scene_range[0]},{scene_range[1]}
+
+        """
+        # run command sh sbatch_file_content
+        with open(f"tmp_sbatch.sh", "w") as f:
+            f.write(sbatch_file_content)
+        print(f"Write sbatch file: tmp_sbatch.sh")
+        os.system(f"sbatch tmp_sbatch.sh")
+
+if __name__ == '__main__':
+    start_time = time.time()
+    fire.Fire(main)
+    print(f"\nTime used: {(time.time() - start_time)/60:.2f} mins")

tests/scene_flow.py

@@ -93,18 +93,18 @@ def vis(
         pose_flow = pc0[:, :3] @ ego_pose[:3, :3].T + ego_pose[:3, 3] - pc0[:, :3]
 
         pcd = o3d.geometry.PointCloud()
-        # pcd.points = o3d.utility.Vector3dVector(pc0[:, :3][~gm0])
-        # pcd.colors = o3d.utility.Vector3dVector(flow_color[~gm0])
-        pcd.points = o3d.utility.Vector3dVector(pc0[:, :3])
         if flow_mode in ['dufo_label', 'label']:
             labels = data[flow_mode]
-            pcd.colors = o3d.utility.Vector3dVector(np.ones_like(pc0[:, :3]))
-            for i in range(pc0.shape[0]):
-                if labels[i] <= 0:
-                    continue
+            pcd_i = o3d.geometry.PointCloud()
+            for label_i in np.unique(labels):
+                pcd_i.points = o3d.utility.Vector3dVector(pc0[labels == label_i][:, :3])
+                if label_i <= 0:
+                    pcd_i.paint_uniform_color([1.0, 1.0, 1.0])
                 else:
-                    pcd.colors[i] = color_map[labels[i] % len(color_map)]
+                    pcd_i.paint_uniform_color(color_map[label_i % len(color_map)])
+                pcd += pcd_i
         elif flow_mode in data:
+            pcd.points = o3d.utility.Vector3dVector(pc0[:, :3])
             flow = data[flow_mode] - pose_flow # ego motion compensation here.
             flow_color = flow_to_rgb(flow) / 255.0
             is_dynamic = np.linalg.norm(flow, axis=1) > 0.1