Skip to content

GETTING_STARTED.md

Latest commit

 

History

History
60 lines (50 loc) · 1.38 KB

GETTING_STARTED.md

File metadata and controls

60 lines (50 loc) · 1.38 KB

Run the training code on the WIBAM dataset

Clone the repository

git clone https://github.com/MatthewHowe/WIBAM.git
cd WIBAM

Create a data directory

mkdir data

Download the WIBAM dataset(alternate link) and organise the directory as follows.

WIBAM
│   README.md
│   requirements.txt    
│   ...
|
└───data
│   └───wibam
|       └───calib
│       └───annotations
│       └───frames
|       |   └───0
|       |   └───1
|       |   └───2
|       |   └───3
│       └───image_sets
|       └───models
│   
└───src
    └───lib
    └───tools
    |   ...

Pull the docker image.

docker pull matthewhowe/wibam

Run the docker image

make run

From the main WIBAM directory run training code

python src/main_lit.py ddd --trainset_percentage=1.0 --output_path= --load_model=models/nuScenes_3Ddetection_e140.pth --dataset=wibam --batch_size=128 --lr=7.8125e-6 --num_workers=10 --gpus=0,1,2,3

From the main WIBAM directory run the test to replicate results

python src/main_lit.py ddd --dataset=wibam --load_model=data/wibam/models/wibam.ckpt --batch_size=1 --save_video --gpus=0 --num_workers=1 --test_only