- Clone the repository
git clone https://github.com/xwaiyy123/object-navigation.git
and move into the top level directorycd object-navigation
- Create conda environment.
pip install -r requirements.txt
- Download the dataset, which refers to ECCV-VN.
- You can also use the DETR object detection features.
python main.py --title IOM --model IOM --workers 12 --gpu-ids 0 1 --max-ep 3000000 --log-dir runs/RL_train --save-model-dir trained_models/RL_train --pretrained-trans trained_models/pretrain/checkpoint0004.pth --data-dir /opt/data/private/datasets/AI2Thor_offline_data_2.0.2/
python full_eval.py --title IOM --model IOM --results-json eval_results/IOM.json --gpu-ids 0 --log-dir runs/RL_train --save-model-dir trained_models/RL --data-dir /opt/data/private/datasets/AI2Thor_offline_data_2.0.2/
You can see the results of each test in 'eval_results/log.txt', and the iteration process of the model in '/result_epoch.txt'. You can download the model weights from this address.
This project is based on the following codebases. -DOA. -ECCV-VN. -VTNet. If you find this project helpful, Please also cite the codebases above. Thanks.