Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/Angzz/fcos-gluon-cv
FCOS: Fully Convolutional One-Stage Object Detection
https://github.com/Angzz/fcos-gluon-cv
computer-vision fcos gluon-cv object-detection
Last synced: about 2 months ago
JSON representation
FCOS: Fully Convolutional One-Stage Object Detection
- Host: GitHub
- URL: https://github.com/Angzz/fcos-gluon-cv
- Owner: Angzz
- License: apache-2.0
- Created: 2019-04-27T15:39:50.000Z (over 5 years ago)
- Default Branch: master
- Last Pushed: 2022-10-09T14:42:13.000Z (almost 2 years ago)
- Last Synced: 2024-06-12T02:52:41.375Z (3 months ago)
- Topics: computer-vision, fcos, gluon-cv, object-detection
- Language: Python
- Size: 1.14 MB
- Stars: 29
- Watchers: 5
- Forks: 7
- Open Issues: 3
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
- Awesome-MXNet - **FCOS**
README
# FCOS: Fully Convolutional One-Stage Object Detection
This is an unofficial implementation of [FCOS](https://arxiv.org/abs/1904.01355) in a [gluon-cv](http://gluon-cv.mxnet.io) style, we implemented this anchor-free framework in a fully [Gluon](https://mxnet.incubator.apache.org/versions/master/gluon/index.html) API, please stay tuned!
## Main Results
| Model | Backbone | Train Size | Batch Size | AP(val) |
| :----------: | :----------: | :----------: | :----------: | :----------: |
| fcos_resnet50_v1_coco | ResNet50-V1 | 800 | 1 | - |
| fcos_resnet50_v1b_coco | ResNet50-V1b | 800 | 1 | 33.1 |
| fcos_resnet101_v1d_coco | ResNet101-V1d | 800 | 1 | 37.5 |
Note: To be update.
## Installation
1. Install cuda `10.0` and mxnet `1.4.0`.
```Shell
sudo pip3 install mxnet-cu100==1.4.0.post0
```
2. Clone the code, and install gluoncv with ``setup.py``.
```Shell
cd fcos-gluon-cv
sudo python3 setup.py build
sudo python3 setup.py install
```
## Preparation
1. Download `COCO2017` datasets follow the official [tutorials](https://gluon-cv.mxnet.io/build/examples_datasets/mscoco.html#sphx-glr-build-examples-datasets-mscoco-py) and create a soft link.
```Shell
ln -s $DOWNLOAD_PATH ~/.mxnet/datasets/coco
```
You can also download from [cocodataset](http://cocodataset.org) and execute the command above.
2. More preparations can also refer to [GluonCV](https://gluon-cv.mxnet.io/index.html).
3. All experiments are performed on `8 * 2080ti` GPU with `Python3.5`, `cuda10.0` and `cudnn7.5.0`.
## Structure
```Shell
* Model : $ROOT/gluoncv/model_zoo/fcos
* Train & valid scripts : $ROOT/scripts/detection/fcos
* Data Transform : $ROOT/gluoncv/data/transform/presets
```
## Training & Inference
1. Clone the training scripts [here](https://github.com/Angzz/fcos-gluon-cv/blob/master/scripts/detection/fcos/train_fcos.py), then train `fcos_resnet50_v1b_coco` with:
```Shell
python3 train_fcos.py --network resnet50_v1b --gpus 0,1,2,3,4,5,6,7 --num-workers 32 --batch-size 8 --log-interval 10
```
2. Clone the eval scripts [here](https://github.com/Angzz/fcos-gluon-cv/blob/master/scripts/detection/fcos/eval_fcos.py), then validate `fcos_resnet50_v1b_coco` with:
```Shell
python3 eval_fcos.py --network resnet50_v1b --gpus 0,1,2,3,4,5,6,7 --num-workers 32 --pretrained $SAVE_PATH/XXX.params
```
## Reference
* **FCOS:** Zhi Tian, Chunhua Shen, Hao Chen, Tong He.
"FCOS: Fully Convolutional One-Stage Object Detection." arXiv (2019). [[paper](https://arxiv.org/pdf/1904.01355)] [[official-code](https://github.com/tianzhi0549/FCOS)] [[mmdet](https://github.com/open-mmlab/mmdetection)]