Skip to content
forked from NVIDIA/DALI

A GPU-accelerated library containing highly optimized building blocks and an execution engine for data processing to accelerate deep learning training and inference applications.

License

Notifications You must be signed in to change notification settings

rsek147/DALI-SSD

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

DALI-SSD

This repository is a forked from NVIDIA/DALI to enable training of various SSD models.

Setup

docker build -t dali-ssd docker
docker run --gpus all -it --name dali-ssd dali-ssd:latest
cd dali-ssd

Download dataset

cd data
bash get_coco.sh

Training

cd docs/examples/use_cases/pytorch/single_stage_detector
python3 -m torch.distributed.launch --nproc_per_node=4 ./main.py --data ../../../../data/coco/ \
                                                                 --epochs 100 \
                                                                 --batch-size 32 \
                                                                 --warmup 300 \
                                                                 --backbone resnet18 \
                                                                 --num-workers 8 \
                                                                 --fp16-mode static

ToDo

  • Backbone
    • MobileNetV2
    • MobileNetV3
    • GhostNet
  • Head
    • SSD Lite
    • SSD 512
  • Data Augmentation
    • Mosaic
    • Cutmix

About

A GPU-accelerated library containing highly optimized building blocks and an execution engine for data processing to accelerate deep learning training and inference applications.

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • C++ 63.9%
  • Python 20.1%
  • Cuda 11.4%
  • CMake 2.3%
  • Shell 1.4%
  • C 0.9%