sporsho / 3D_OutDet

MIT License
7 stars 0 forks source link

Need Requirement #1

Closed qq625924821 closed 2 months ago

qq625924821 commented 2 months ago

Hello, Could you provide the requirement for me? And tell me the version of python. I want to use your code for filtering snow! Thanck you very much!

sporsho commented 2 months ago

Hi, You can use this file as a requirement file. If conda fails to install all the libraries, use pip to install the missing libraries. https://github.com/sporsho/3D_OutDet/blob/main/wads_8_sep_2023.yml

Alternatively, you can also use the below shell script to create the environment. This is a minimal environment, not all the functionalities may be available with it. You can also update the script with any compatible library versions.

#! /bin/bash 
echo "Conda version: "
conda -V 
echo "Attempting to update conda"
yes | conda update conda

echo "Creating environment"
yes | conda create -n outdet python=3.10
if [ $? -eq 0 ]; then
    echo "basic invironment installed"
    echo "Assuming you are using Bash"
    BASE=$(conda info --base)
    source $BASE/etc/profile.d/conda.sh
    conda activate outdet 
    pip config set global.extra-index-url https://pypi.ngc.nvidia.com
    yes | conda install numpy=1.23.1
    yes | conda install cudatoolkit=11.3.1
    yes | conda install pytorch=1.12.1=py3.10_cuda11.3_cudnn8.3.2_0 -c pytorch
    yes | pip install pyyaml==6.0
    yes | pip install tqdm==4.46.1
    yes | pip install scikit-image==0.19.3
    yes | pip install scikit-learn==1.1.2
    yes | pip install scipy==1.9.1
    yes | pip install matplotlib==3.6.0
    yes | pip install cubinlinker-cu11==0.3.0.post1 --extra-index-url https://pypi.nvidia.com
    yes | pip install cuda-python==11.8.1
    yes | pip install cudf-cu11==23.6.0 --extra-index-url https://pypi.nvidia.com
    yes | pip install cuml-cu11==23.6.0 --extra-index-url https://pypi.nvidia.com
    yes | pip install cupy-cuda11x==12.1.0 
    yes | pip install torch-cluster==1.6.0
    which python 
    conda deactivate
else    
    echo "environment installation failed"
fi
qq625924821 commented 1 month ago

thank you very much!

---Original--- From: @.> Date: Mon, Jul 22, 2024 16:45 PM To: @.>; Cc: @.**@.>; Subject: Re: [sporsho/3D_OutDet] Need Requirement (Issue #1)

Hi, You can use this file as a requirement file. If conda fails to install all the libraries, use pip to install the missing libraries. https://github.com/sporsho/3D_OutDet/blob/main/wads_8_sep_2023.yml

Alternatively, you can also use the below shell script to create the environment. This is a minimal environment, not all the functionalities may be available with it. You can also update the script with any compatible library versions.

! /bin/bash echo "Conda version: " conda -V echo "Attempting to update conda" yes | conda update conda echo "Creating environment" yes | conda create -n outdet python=3.10 if [ $? -eq 0 ]; then echo "basic invironment installed" echo "Assuming you are using Bash" BASE=$(conda info --base) source $BASE/etc/profile.d/conda.sh conda activate outdet pip config set global.extra-index-url https://pypi.ngc.nvidia.com yes | conda install numpy=1.23.1 yes | conda install cudatoolkit=11.3.1 yes | conda install pytorch=1.12.1=py3.10_cuda11.3_cudnn8.3.2_0 -c pytorch yes | pip install pyyaml==6.0 yes | pip install tqdm==4.46.1 yes | pip install scikit-image==0.19.3 yes | pip install scikit-learn==1.1.2 yes | pip install scipy==1.9.1 yes | pip install matplotlib==3.6.0 yes | pip install cubinlinker-cu11==0.3.0.post1 --extra-index-url https://pypi.nvidia.com yes | pip install cuda-python==11.8.1 yes | pip install cudf-cu11==23.6.0 --extra-index-url https://pypi.nvidia.com yes | pip install cuml-cu11==23.6.0 --extra-index-url https://pypi.nvidia.com yes | pip install cupy-cuda11x==12.1.0 yes | pip install torch-cluster==1.6.0 which python conda deactivate else echo "environment installation failed" fi

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

qq625924821 commented 1 month ago

Hi, there are three pt models in the 3D_OutDet project, such as kitti、spray、wads. Can these models be used with each other?

------------------ 原始邮件 ------------------ 发件人: "sporsho/3D_OutDet" @.>; 发送时间: 2024年7月22日(星期一) 下午4:45 @.>; @.**@.>; 主题: Re: [sporsho/3D_OutDet] Need Requirement (Issue #1)

Hi, You can use this file as a requirement file. If conda fails to install all the libraries, use pip to install the missing libraries. https://github.com/sporsho/3D_OutDet/blob/main/wads_8_sep_2023.yml

Alternatively, you can also use the below shell script to create the environment. This is a minimal environment, not all the functionalities may be available with it. You can also update the script with any compatible library versions.

! /bin/bash echo "Conda version: " conda -V echo "Attempting to update conda" yes | conda update conda echo "Creating environment" yes | conda create -n outdet python=3.10 if [ $? -eq 0 ]; then echo "basic invironment installed" echo "Assuming you are using Bash" BASE=$(conda info --base) source $BASE/etc/profile.d/conda.sh conda activate outdet pip config set global.extra-index-url https://pypi.ngc.nvidia.com yes | conda install numpy=1.23.1 yes | conda install cudatoolkit=11.3.1 yes | conda install pytorch=1.12.1=py3.10_cuda11.3_cudnn8.3.2_0 -c pytorch yes | pip install pyyaml==6.0 yes | pip install tqdm==4.46.1 yes | pip install scikit-image==0.19.3 yes | pip install scikit-learn==1.1.2 yes | pip install scipy==1.9.1 yes | pip install matplotlib==3.6.0 yes | pip install cubinlinker-cu11==0.3.0.post1 --extra-index-url https://pypi.nvidia.com yes | pip install cuda-python==11.8.1 yes | pip install cudf-cu11==23.6.0 --extra-index-url https://pypi.nvidia.com yes | pip install cuml-cu11==23.6.0 --extra-index-url https://pypi.nvidia.com yes | pip install cupy-cuda11x==12.1.0 yes | pip install torch-cluster==1.6.0 which python conda deactivate else echo "environment installation failed" fi

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

sporsho commented 1 month ago

No, there are multiple domain shifts between the datasets hence if you want to use a model trained using wads to infer on snowykitti dataset, you have to apply some fine-tuning with some domain transfer technique. Alternatively, you can do a generalized training. e..g, merge all the training data from snowykitti, wads and spray and train a generalized model. The performance will decrease, but you can apply the learned model on any of the datasets.

qq625924821 commented 1 month ago

Are there some domain transfer example in you 3D_OutDet project? or could you recommend some relevant materials to me? Thank you very much!

------------------ 原始邮件 ------------------ 发件人: "sporsho/3D_OutDet" @.>; 发送时间: 2024年8月5日(星期一) 下午3:23 @.>; @.**@.>; 主题: Re: [sporsho/3D_OutDet] Need Requirement (Issue #1)

No, there are multiple domain shifts between the datasets hence if you want to use a model trained using wads to infer on snowykitti dataset, you have to apply some fine-tuning with some domain transfer technique. Alternatively, you can do a generalized training. e..g, merge all the training data from snowykitti, wads and spray and train a generalized model. The performance will decrease, but you can apply the learned model on any of the datasets.

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

sporsho commented 1 month ago

No, our project is not about domain transfer or adaptation. It is another stream of research and though I did my masters thesis on semi-supervised domain adaptation for MRI images, at present I am not involved with any kind of domain adaptation yet. For starter, you can search on the google scholar e.g., https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=domain+adaptation+point+cloud+segmentation&btnG=

qq625924821 commented 1 month ago

OK,Thanks! I will try to train it on my datasets.

------------------ 原始邮件 ------------------ 发件人: "sporsho/3D_OutDet" @.>; 发送时间: 2024年8月5日(星期一) 下午3:23 @.>; @.**@.>; 主题: Re: [sporsho/3D_OutDet] Need Requirement (Issue #1)

No, there are multiple domain shifts between the datasets hence if you want to use a model trained using wads to infer on snowykitti dataset, you have to apply some fine-tuning with some domain transfer technique. Alternatively, you can do a generalized training. e..g, merge all the training data from snowykitti, wads and spray and train a generalized model. The performance will decrease, but you can apply the learned model on any of the datasets.

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>