ResNet18--Image-Classification and ResNet50--Object Detection

Checkpoints, logs and source code

Dependencies

本实验在Ubuntu20.04，CUDA11.8上成功运行代码

• Python=3.10
• PyTorch=2.4.1
• 其他依赖直接pip install 即可

Usage

Swanlab

This project use Swanlab to log information and report. Remember to adjust the code in main.py to suit your research.

Training

可以直接终端运行train_test.sh文件：./train_test.sh

CUDA_VISIBLE_DEVICES=0 OMP_NUM_THREADS=1 torchrun --nnodes 1 --nproc_per_node 1 --master_port 49945  main.py \
--cfg 'configs/ResNet/ResNet18.yaml' \
--train_data_path 'data/train' \
--test_data_path 'data/val' \
--output 'Output' \
--tag 'Image_Class_test_Top_1_model' \
--Top_K 1\

CUDA_VISIBLE_DEVICES=0,1,2,3 OMP_NUM_THREADS=1 torchrun --nnodes 1 --nproc_per_node 4 --master_port 49945  VOC_Object_Detection.py \
--cfg 'configs/ResNet/ResNet50.yaml' \
--train_data_path 'Data' \
--test_data_path 'Data' \
--output 'Output' \
--tag 'Object_Detection'\

Acknowledgement

We borrowed some parts from the following open-source projects:

• DEIQT