CSNet

model: https://pan.baidu.com/s/1pedD8ebX2F1d4dcXeFHZbg code: ed4i

This model is not the optimal sampling model; however, training in accordance with the description in the paper yields the optimal model.

==============================

1. Requirements ==============================

• OS: Ubuntu 20.04 LTS
• Python: 3.8 ~ 3.11
• CUDA Toolkit: 11.8
• Build tools: gcc/g++
• PyTorch: >= 2.0

============================== 2) Environment Setup

Update system

sudo apt update && sudo apt upgrade -y

Install dependencies

sudo apt install build-essential git python3-dev python3-pip -y

Create virtual environment (optional)

python3 -m venv csnet-env source csnet-env/bin/activate

Install PyTorch

pip install --index-url https://download.pytorch.org/whl/cu118 torch torchvision torchaudio

Install other dependencies

pip install numpy tqdm scipy tensorboard open3d

============================== 3) Dataset Preparation

Example structure: CSNet/ ├── command_files/ │ ├── train_ds_model.py │ └── test_ds_model.py ├── data/ │ └── output1024/ │ ├── train/ # training .ply files │ └── test/ # testing .ply files └── save_model/ # directory for checkpoints

============================== 4) Training Command

Run from repo root

python command_files/train_ds_model.py
--train_dir data/output1024/train
--test_dir data/output1024/test
--batch_size 32
--epochs 200
--lr 1e-3
--save_dir save_model/best_512_cash.mdl

============================== 5) Testing Command

Run evaluation with trained checkpoint

python command_files/test_ds_model.py
--test_dir data/output1024/test
--ckpt save_model/best_512_cash.mdl
--batch_size 32