Skip to content

Latest commit

 

History

History
65 lines (48 loc) · 1.8 KB

INSTALL.md

File metadata and controls

65 lines (48 loc) · 1.8 KB

Installation

Requirements

  • Linux (Windows is not officially supported)
  • Python 3.7
  • PyTorch 1.6.0 & TorchVision 0.7.0
  • CUDA 10.1
  • mmcv-full 1.3.12

Install MMHOIDet

a. Create a conda (recommended) virtual environment and activate it.

conda create -n mmhoidet python=3.7 -y
conda activate mmhoidet

b. Install PyTorch and TorchVision following the official instructions, e.g.,

conda install pytorch==1.6.0 torchvision==0.7.0 cudatoolkit=10.1 -c pytorch -y

c. Clone the MMHOIDet repository.

git clone https://github.com/noobying/mmhoidet
cd mmhoidet

d. Install full dependencies and package.

pip install -e .[full]

Prepare datasets

It is recommended to symlink the dataset root to mmhoidet/data. If your folder structure is different, you may need to change the corresponding paths in config files.

mmhoidet
├── mmhoidet
├── tools
├── configs
├── data
│   ├── hico_20160224_det
│   │   ├── annotations
│   │   |   ├── test_hico.json
│   │   |   ├── trainval_hico.json
│   │   ├── images
│   │   |   ├── test2015
│   │   |   ├── train2015

note: HICO-Det can be downloaded here. The test_hico.json and trainval_hico.json in annotations is provided by PPDM. You can download the annotations from here and replace the original annotations directory.

Besides, you need to do some tiny changes to the dataset annotation file with the following command:

python ./tools/data_converters/hico_det.py ./data/hico_20160224_det --data_type test train