The code is tested with PyTorch 1.10.0 and CUDA 11.3. After cloning the repository, follow the below steps for installation,
- Python ≥ 3.7
- PyTorch ≥ 1.8
- Detectron2: follow Detectron2 installation instructions.
Our project uses a submodule mavl for generating psuedo proposals from MViT model MAVL. The proposals have been
provided with the datasets and detailed in DATASETS.md. The submodule is required to generate the pseudo-proposals.
If you forget to add --recurse-submodules when you clone the repository, do:
- Update the submodules [optional]
git clone https://github.com/hanoonaR/object-centric-ovd.git
git submodule init
git submodule update- Install PyTorch and torchvision
Installation with pip:
pip install torch==1.10.0 torchvision==0.11.0 torchaudio --extra-index-url https://download.pytorch.org/whl/cu113or install with conda:
conda install pytorch==1.10.0 torchvision==0.11.0 torchaudio==0.10.0 cudatoolkit=11.3 -c pytorch -c conda-forge- Install Detectron2 Install with pre-build Detectron2:
python -m pip install detectron2 -f \
https://dl.fbaipublicfiles.com/detectron2/wheels/cu113/torch1.10/index.htmlor install from source:
git clone [email protected]:facebookresearch/detectron2.git
cd detectron2
pip install -e .- Install other dependencies
pip install -r requirements.txt