空间站广场

论文

Notebooks

比赛

课程

Apps

我的主页

我的Notebooks

我的论文库

我的足迹

我的工作空间

任务

节点

文件

数据集

镜像

项目

数据库

公开

LimitIRSTD-Track1 Submission Example

AI4SCUP-PRCV

赛事组委会-应昕怡

更新于 2024-06-22

推荐镜像 :prcv-infer:1.6

推荐机型 :c8_m31_1 * NVIDIA T4

数据集

LimitIRSTD-Track1示例数据(v1)

Submission Example

代码

文本

alt

Click the "New dataset" button to import "Local Code".
Click the "Import" button to import the your personal notebook. See the code submission example below.

代码

文本

Submission Example (SIRST3 Example dataset)

For personal notebook code debugging

代码

文本

[1]

# Download the LESPS code to the evaluation platform locally

# Participants choose to download their own github code or upload their own local code via the "New Dataset" button.

# Note: The file uploaded through the "New dataset" button only has read permission but not write permission, so you need to copy it through the cp command.

!rm -rf /LESPS

!git clone https://github.com/XinyiYing/LESPS

代码

文本

[2]

# Load the example dataset: SIRST3

!mkdir -p /LESPS/datasets/SIRST3 # Create a new folder to store the example dataset

!cp -r /bohr/LimitIRSTD-Track1-Sample-7mq8/v1/* /LESPS/datasets/SIRST3 # Download the sample data to /LESPS/dataset/SIRST3

# Load the test dataset

# Note that, when participants submit their code, the path of the test set is $DATA_PATH/*, and the data set name can be customized.

# !mkdir -p /LESPS/datasets/LimitIRSTD # Create a new folder to store the test data

# !cp -r $DATA_PATH/* /LESPS/datasets/LimitIRSTD # Download the test dataset to /LESPS datasets/LimitIRSTD

代码

文本

[7]

%%bash

# Complete the "training process" of DNANet_LESPS_coarse model on SIRST3 under coarse point supervision.

# To simplify the process, only 2 epochs are provided, the default setting is 400 epochs.

cd /LESPS

python train.py --model_names DNANet --dataset_names SIRST3 --label_type 'coarse' --nEpochs 2

代码

文本

[8]

%%bash

# Complete the "training process" of DNANet model on SIRST3 under full supervision.

# To simplify the process, only 2 epochs are provided, the default setting is 400 epochs.

cd /LESPS

python train_full.py --model_names DNANet --dataset_names SIRST3 --label_type 'full' --nEpochs 2

代码

文本

[10]

%%bash

# Complete the "inference and evaluation process" of DNANet_LESPS_coarse model on SIRST3.

cd /LESPS

python test.py --model_names DNANet --dataset_names SIRST3 --pth_dirs SIRST3/DNANet_LESPS_coarse.pth.tar

代码

文本

[6]

%%bash

# Complete the "inference process" of DNANet_LESPS_coarse model on SIRST3.

# Results are saved in /LESPS/results/SIRST3/DNANet_LESPS_coarse/

cd /LESPS

python inference.py --model_names DNANet --dataset_names SIRST3 --pth_dirs SIRST3/DNANet_LESPS_coarse.pth.tar --save_img True

代码

文本

[11]

%%bash

# Compress the results into submission.zip, and move it to the specified folder.

# Note that: The ranking process is automatically running.

original_dir=$PWD

cd /LESPS/results/SIRST3/ # cd path

cp -r DNANet_LESPS_coarse mask

zip -r submission.zip mask/ # Compress submission.zip

mv submission.zip $original_dir/ # Move submission.zip to the specified folder for automatically ranking.

代码

文本

Submission sample (Test dataset)

For code submission and leaderboard scoring

代码

文本

[ ]

%%bash

# Download the LESPS code

rm -rf LESPS

git clone https://github.com/XinyiYing/LESPS

代码

文本

[ ]

%%bash

# Load the test dataset

mkdir -p LESPS/datasets/LimitIRSTD

cp -r $DATA_PATH/* LESPS/datasets/LimitIRSTD

cp -r LESPS/datasets/LimitIRSTD/img_idx/test.txt LESPS/datasets/LimitIRSTD/img_idx/test_LimitIRSTD.txt

cp -r LESPS/datasets/LimitIRSTD/img_idx/test.txt LESPS/datasets/LimitIRSTD/img_idx/train_LimitIRSTD.txt

代码

文本

[ ]

%%bash

# Download the weight baseline model

cd LESPS

wget https://bohrium.oss-cn-zhangjiakou.aliyuncs.com/competition/15/DNANet_LESPS_coarse_400.pth.tar

mkdir -p log/LimitIRSTD/

mv DNANet_LESPS_coarse_400.pth.tar log/LimitIRSTD/

代码

文本

[ ]

# Complete inference process，and save the results in LESPS/results/LimitIRSTD/DNANet_LESPS_coarse/

!cd LESPS && python inference.py --model_names DNANet --dataset_names LimitIRSTD --pth_dirs LimitIRSTD/DNANet_LESPS_coarse_400.pth.tar --img_norm_cfg_mean 76.048 --img_norm_cfg_std 28.732

代码

文本

[ ]

%%bash

# Compress the results into submission.zip, and move it to the specified folder.

# Note that: The ranking process is automatically running.

original_dir=$PWD

cd LESPS/results/LimitIRSTD/ # cd path

cp -r DNANet_LESPS_coarse_400 mask

zip -r submission.zip mask/ # Compress submission.zip

mv submission.zip $original_dir/ # Move submission.zip to the specified folder for automatically ranking.

代码

文本

Directory of test dataset

├──images
│    ├── 00001.png
│    ├── 00002.png
│    ├── ...
│    ├── 00500.png
├── img_idx
│    ├── test.txt
├── statistics.txt

代码

文本

AI4SCUP-PRCV

点个赞吧