Skip to content

Latest commit

 

History

History
63 lines (51 loc) · 2.31 KB

GETTING_STARTED.md

File metadata and controls

63 lines (51 loc) · 2.31 KB
Raw

Getting Started

Our code is tested on Ubuntu. I have briefly tested the GUI on Windows. Might also work on MAC (#103) but I haven't tested.

Requirements

  • Python 3.8+
  • PyTorch 1.11+ (See PyTorch for installation instructions)
  • torchvision corresponding to the PyTorch version
  • OpenCV (try pip install opencv-python)
  • Others: pip install -r requirements.txt

Dataset

I recommend either softlinking (ln -s) existing data or use the provided scripts/download_datasets.py to structure the datasets as our format.

python -m scripts.download_datasets

The structure is the same as the one in STCN -- you can place XMem in the same folder as STCN and it will work. The script uses Google Drive and sometimes fails when certain files are blocked from automatic download. You would have to do some manual work in that case. It does not download BL30K because it is huge and we don't want to crash your harddisks.

├── XMem
├── BL30K
├── DAVIS
│   ├── 2016
│   │   ├── Annotations
│   │   └── ...
│   └── 2017
│       ├── test-dev
│       │   ├── Annotations
│       │   └── ...
│       └── trainval
│           ├── Annotations
│           └── ...
├── static
│   ├── BIG_small
│   └── ...
├── long_video_set
│   ├── long_video
│   ├── long_video_x3
│   ├── long_video_davis
│   └── ...
├── YouTube
│   ├── all_frames
│   │   └── valid_all_frames
│   ├── train
│   ├── train_480p
│   └── valid
└── YouTube2018
    ├── all_frames
    │   └── valid_all_frames
    └── valid

Long-Time Video

It comes from AFB-URR. Please following their license when using this data. We release our extended version (X3) and corresponding _davis versions such that the DAVIS evaluation can be used directly. They can be downloaded [here]. The script above would also attempt to download it. Use [this script] to expand it to (X3).

BL30K

Download via https://doi.org/10.13012/B2IDB-1702934_V1