angzz / fcos-gluon-cv

FCOS: Fully Convolutional One-Stage Object Detection

FCOS: Fully Convolutional One-Stage Object Detection

This is an unofficial implementation of FCOS in a gluon-cv style, we implemented this anchor-free framework in a fully Gluon 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.

sudo pip3 install mxnet-cu100==1.4.0.post0

2. Clone the code, and install gluoncv with setup.py.

cd fcos-gluon-cv
sudo python3 setup.py build
sudo python3 setup.py install

Preparation

1. Download COCO2017 datasets follow the official tutorials and create a soft link.

ln -s $DOWNLOAD_PATH ~/.mxnet/datasets/coco

You can also download from cocodataset and execute the command above.

2. More preparations can also refer to GluonCV.

3. All experiments are performed on 8 * 2080ti GPU with Python3.5, cuda10.0 and cudnn7.5.0.

Structure

* 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, then train fcos_resnet50_v1b_coco with:

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, then validate fcos_resnet50_v1b_coco with:

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] [official-code] [mmdet]

fcos-gluon-cv's Issues

Would you mind uploading these three training models.

It'll be convinient to run some simple tests. Thanks

Getting Zeros from target generator for cls_targets, ctr_targets and box_targets

[0. 0. 0. ... 0. 0. 0.] <NDArray 20938 @cpu(0)> [0. 0. 0. ... 0. 0. 0.] <NDArray 20938 @cpu(0)> [[0. 0. 0. 0.] [0. 0. 0. 0.] [0. 0. 0. 0.] ... [0. 0. 0. 0.] [0. 0. 0. 0.] [0. 0. 0. 0.]] <NDArray 20938x4 @cpu(0)> [[ 0. 0.] [128. 0.] [256. 0.] ... [976. 992.] [984. 992.] [992. 992.]] <NDArray 20938x2 @cpu(0)>
I'm getting this from the target generator for cls_targets, ctr_targets, box_targets, cor_targets. My input image is (1000, 1000, 3) and the boxes are [291, 114, 901, 778, 60], [504, 263, 780, 490, 15], [461, 222, 829, 579, 20], [ 24, 205, 389, 800, 15]
What is the possible cause for this?
I'm using resnet50 as the base.

It's too slow to train moudle.

Hello， 实验结果出来了吗？