Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/NVlabs/FreeSOLO
FreeSOLO for unsupervised instance segmentation, CVPR 2022
https://github.com/NVlabs/FreeSOLO
cvpr2022 freesolo instance-segmentation object-detection self-supervised-learning
Last synced: about 1 month ago
JSON representation
FreeSOLO for unsupervised instance segmentation, CVPR 2022
- Host: GitHub
- URL: https://github.com/NVlabs/FreeSOLO
- Owner: NVlabs
- License: other
- Created: 2022-03-27T22:33:03.000Z (over 2 years ago)
- Default Branch: main
- Last Pushed: 2023-01-16T08:24:53.000Z (over 1 year ago)
- Last Synced: 2024-04-20T05:35:02.728Z (5 months ago)
- Topics: cvpr2022, freesolo, instance-segmentation, object-detection, self-supervised-learning
- Language: Python
- Homepage:
- Size: 2.2 MB
- Stars: 309
- Watchers: 4
- Forks: 30
- Open Issues: 13
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# FreeSOLO: Learning to Segment Objects without Annotations
This project hosts the code for implementing the FreeSOLO algorithm for unsupervised instance segmentation.
> [**FreeSOLO: Learning to Segment Objects without Annotations**](https://arxiv.org/abs/2202.12181),
> Xinlong Wang, Zhiding Yu, Shalini De Mello, Jan Kautz, Anima Anandkumar, Chunhua Shen, Jose M. Alvarez
> In: Proc. IEEE Conf. Computer Vision and Pattern Recognition (CVPR), 2022
> *arXiv preprint ([arXiv 2202.12181](https://arxiv.org/abs/2202.12181))*
## Visual Results
## Installation
### Prerequisites
- Linux or macOS with Python >= 3.6
- PyTorch >= 1.5 and torchvision that matches the PyTorch installation.
- scikit-image
### Install PyTorch in Conda env
```shell
# create conda env
conda create -n detectron2 python=3.6
# activate the enviorment
conda activate detectron2
# install PyTorch >=1.5 with GPU
conda install pytorch torchvision -c pytorch
```
### Build Detectron2 from Source
Follow the [INSTALL.md](https://github.com/facebookresearch/detectron2/blob/master/INSTALL.md) to install Detectron2 (commit id [11528ce](https://github.com/facebookresearch/detectron2/tree/11528ce083dc9ff83ee3a8f9086a1ef54d2a402f) has been tested).
### Datasets
Follow the [datasets/README.md](https://github.com/facebookresearch/detectron2/blob/master/datasets/README.md) to set up the MS COCO dataset.
### Pre-trained model
Download the [DenseCL](https://github.com/WXinlong/DenseCL) pre-trained model from [here](https://cloudstor.aarnet.edu.au/plus/s/4sugyvuBOiMXXnC/download). Convert it to detectron2's format and put the converted model under "training_dir/pre-trained/DenseCL" directory.
```shell
python tools/convert-pretrain-to-detectron2.py {WEIGHT_FILE}.pth {WEIGHT_FILE}.pkl
```
## Usage
### Free Mask
Download the prepared free masks in json format from [here](https://cloudstor.aarnet.edu.au/plus/s/AGcBm8cbqTWEKRS/download). Put it under "datasets/coco/annotations" directory. Or, generate it by yourself:
```shell
bash inference_freemask.sh
```
### Training
```shell
# train with free masks
bash train.sh
# generate pseudo labels
bash gen_pseudo_labels.sh
# self-train
bash train_pl.sh
```
### Testing
Download the trained model from [here](https://cloudstor.aarnet.edu.au/plus/s/V8C0onE5H63x3RD/download).
```shell
bash test.sh {MODEL_PATH}
```
Disclaimer: This is a reimplementation and the numbers are slightly different from the original paper.
The above script will get instance segmentation results of 4.3 AP and 9.4 AP50.
## Citations
Please consider citing our paper in your publications if the project helps your research. BibTeX reference is as follow.
```BibTeX
@article{wang2022freesolo,
title={{FreeSOLO}: Learning to Segment Objects without Annotations},
author={Wang, Xinlong and Yu, Zhiding and De Mello, Shalini and Kautz, Jan and Anandkumar, Anima and Shen, Chunhua and Alvarez, Jose M},
journal={arXiv preprint arXiv:2202.12181},
year={2022}
}
```