Mask TextSpotter v3
号称最强端到端文本识别模型
实际测试,检测效果确实非同凡响,速度上稍微差于ABCNet.先上图看看效果
同样存在部分漏检情况。但是第一张图,这个居然检测出来了,还是很强大的,要知道ABCNet对第一张图里的,检测实在太差。
速度上,如下,
ABC:
mst:
环境及脚本如下:
Python3 (Python3.7 is recommended)
PyTorch >= 1.4 (1.4 is recommended)
cocoapi
yacs
matplotlib
GCC >= 4.9 (This is very important!)
OpenCV
CUDA >= 9.0 (10.0.130 is recommended)
my env:
torch 1.4.0
torchvision 0.5.0
py362
python 3.6.2
cuda 10.1
git clone https://github.com/NVIDIA/apex.git
cd apex
python setup.py install --cuda_ext --cpp_ext
# 注意:apex编译在torch==1.5.1 torchvision==0.6.1下的
cd MaskTextSpotterV3
# build
python setup.py build develop
# demo :a single image inference by python tools/demo.py
eg:
# 1.single image:
python tools/demo.py --image_path ./demo_images/img_77.jpg --visu_path ./demo_images/img_77_res.jpg
# 2.image folder
eg:
python tools/demo.py --input ./demo_images/ --output ./out/demo_out
# test
python tools/test_net.py --config-file configs/mixtrain/seg_rec_poly_fuse_feature.yaml
# train
# 1.Trained with SynthText
python3 -m torch.distributed.launch --nproc_per_node=8 tools/train_net.py --config-file configs/pretrain/seg_rec_poly_fuse_feature.yaml
训练部分,暂时未进行。