# ovsam
**Repository Path**: data_factory/ovsam
## Basic Information
- **Project Name**: ovsam
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: BSD-3-Clause
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No
## Statistics
- **Stars**: 0
- **Forks**: 0
- **Created**: 2024-05-11
- **Last Updated**: 2024-05-30
## Categories & Tags
**Categories**: Uncategorized
**Tags**: None
## README
# Open-Vocabulary SAM
[Haobo Yuan1](https://yuanhaobo.me),
[Xiangtai Li1](https://lxtgh.github.io),
[Chong Zhou1](https://chongzhou96.github.io),
[Yining Li2](https://scholar.google.com/citations?user=y_cp1sUAAAAJ),
[Kai Chen2](https://chenkai.site),
[Chen Change Loy1](https://www.mmlab-ntu.com/person/ccloy/).
[1S-Lab, Nanyang Technological University](https://www.mmlab-ntu.com/),
[2Shanghai Artificial Intelligence Laboratory](https://www.shlab.org.cn/)
[](https://arxiv.org/abs/2401.02955)
[](https://www.mmlab-ntu.com/project/ovsam)
[](https://huggingface.co/spaces/HarborYuan/ovsam)
[](https://openxlab.org.cn/apps/detail/houshaowei/Open-Vocabulary_SAM)
## π Overview
We introduce the Open-Vocabulary SAM, a SAM-inspired model designed for simultaneous interactive segmentation and recognition, leveraging two unique knowledge transfer modules: SAM2CLIP and CLIP2SAM. The former adapts SAM's knowledge into the CLIP via distillation and learnable transformer adapters, while the latter transfers CLIP knowledge into SAM, enhancing its recognition capabilities.
## π§Usage
To play with Open-Vocabulary SAM, you can:
1. Try the online demo on the [π€Hugging Face Space](https://huggingface.co/spaces/HarborYuan/ovsam). Thanks for the generous support of the Hugging Face team.
2. Run the gradio demo locally by cloning and running the [repo](https://huggingface.co/spaces/HarborYuan/ovsam/tree/main) on π€Hugging Face:
```commandline
git lfs install
git clone https://huggingface.co/spaces/HarborYuan/ovsam ovsam_demo
cd ovsam_demo
conda create -n ovsam_demo python=3.10 && conda activate ovsam_demo
python -m pip install gradio==4.7.1
python -m pip install -r requirements.txt
python main.py
```
3. Try to train or evaluate in this repo following the instructions below.
## βοΈ Installation
We use conda to manage the environment.
Pytorch installation:
```commandline
conda install pytorch torchvision torchaudio cuda-toolkit pytorch-cuda==12.1 -c pytorch -c "nvidia/label/cuda-12.1.0"
```
mmengine installation:
```commandline
python -m pip install https://github.com/open-mmlab/mmengine/archive/refs/tags/v0.8.5.zip
```
mmcv installation (note that older version mmcv before this commit may cause bugs):
```commandline
TORCH_CUDA_ARCH_LIST="{COMCAP}" TORCH_NVCC_FLAGS="-Xfatbin -compress-all" CUDA_HOME=$(dirname $(dirname $(which nvcc))) LD_LIBRARY_PATH=$(dirname $(dirname $(which nvcc)))/lib MMCV_WITH_OPS=1 FORCE_CUDA=1 python -m pip install git+https://github.com/open-mmlab/mmcv.git@4f65f91db6502d990ce2ee5de0337441fb69dd10
```
Please ask ChatGPT to get `COMCAP`:
```text
What is the `Compute Capability` of NVIDIA {YOUR GPU MODEL}? Please only output the number, without text.
```
Other OpenMMLab packages:
```commandline
python -m pip install \
https://github.com/open-mmlab/mmdetection/archive/refs/tags/v3.1.0.zip \
https://github.com/open-mmlab/mmsegmentation/archive/refs/tags/v1.1.1.zip \
https://github.com/open-mmlab/mmpretrain/archive/refs/tags/v1.0.1.zip
```
Extra packages:
```commandline
python -m pip install git+https://github.com/cocodataset/panopticapi.git \
git+https://github.com/HarborYuan/lvis-api.git \
tqdm terminaltables pycocotools scipy tqdm ftfy regex timm scikit-image kornia
```
## π Datasets
Datasets should be put in the `data/` folder of this project similar to [mmdet](https://mmdetection.readthedocs.io/en/latest/user_guides/tracking_dataset_prepare.html). Please prepare dataset in the following format.
### COCO dataset
```text
βββ coco
β βββ annotations
β β βββ panoptic_{train,val}2017.json
β β βββ instance_{train,val}2017.json
β βββ train2017
β βββ val2017
β βββ panoptic_{train,val}2017/ # png annotations
```
### SAM dataset
```text
βββ sam
β βββ train.txt
β βββ val.txt
β βββ sa_000020
β β βββ sa_223750.jpg
β β βββ sa_223750.json
β β βββ ...
β βββ ...
```
`train.txt` and `val.txt` should contain all the folders you need:
```text
sa_000020
sa_000021
...
```
## π Training
Please extract the language embeddings first.
```commandline
bash tools/dist.sh gen_cls seg/configs/ovsam/ovsam_coco_rn50x16_point.py 8
```
### SAM2CLIP
SAM feature extraction:
```commandline
bash tools/dist.sh test seg/configs/sam2clip/sam_vith_dump.py 8
```
SAM2CLIP training:
```commandline
bash tools/dist.sh train seg/configs/sam2clip/sam2clip_vith_rn50x16.py 8
```
### CLIP2SAM
CLIP2SAM training:
```commandline
bash tools/dist.sh train seg/configs/clip2sam/clip2sam_coco_rn50x16.py 8
```
## πββοΈInference
```commandline
bash tools/dist.sh test seg/configs/ovsam/ovsam_coco_rn50x16_point.py 8
```
Please refer to [π€Hugging Face](https://huggingface.co/HarborYuan/ovsam_models) to get the pre-trained weights:
```commandline
git clone https://huggingface.co/HarborYuan/ovsam_models models
```
## π Citation
```bibtex
@article{yuan2024ovsam,
title={Open-Vocabulary SAM: Segment and Recognize Twenty-thousand Classes Interactively},
author={Yuan, Haobo and Li, Xiangtai and Zhou, Chong and Li, Yining and Chen, Kai and Loy, Chen Change},
journal={arXiv preprint},
year={2024}
}
```
## License
This project is licensed under NTU S-Lab License 1.0. Redistribution and use should follow this license.