search

princeton-vl / CornerNet

BSD 3-Clause "New" or "Revised" License
2.36k stars 475 forks source link

how to use the voc to train #55

Open maggiequan opened 5 years ago

maggiequan commented 5 years ago

hello, i am using the voc datasets to train. now, i have the xml、jpg and trainval.txt, i have no idea what i need to do next.convert the xml to json and the use the json to generate the trainval.json?or convert the trainval.txt to trainval.json? i don't know how to solve the problem, can you tell me what to do next?thank you!

zhaowujie commented 5 years ago

voc转coco的matlab代码 clear all; mex('CXXFLAGS=\$CXXFLAGS -std=c++11 -Wall','-largeArrayDims',...
'private/gasonMex.cpp','../common/gason.cpp',...
'-I../common/','-outdir','private');
% 在cocoapi/matlabAPI下运行，matlabAPI/cocoytils.m line54加上文件的名字,检查哪个xml文件出问题 % 在matlabAPI/cocoytils.m line47 修改图像文件的扩展名 % 在VOCcode/VOCinit.m line 81 修改类别的名字 CocoUtils.convertPascalGt( 'E:/data/jinju', '2007', 'val', 'E:/data/jinju/pascal_val2007.json')
CocoUtils.convertPascalGt( 'E:/data/jinju', '2007', 'trainval', 'E:/data/jinju/pascal_train2007.json') @maggiequan

maggiequan commented 5 years ago

voc转coco的matlab代码 clear all; mex('CXXFLAGS=$CXXFLAGS -std=c++11 -Wall','-largeArrayDims',... 'private/gasonMex.cpp','../common/gason.cpp',... '-I../common/','-outdir','private'); % 在cocoapi/matlabAPI下运行，matlabAPI/cocoytils.m line54加上文件的名字,检查哪个xml文件出问题 % 在matlabAPI/cocoytils.m line47 修改图像文件的扩展名 % 在VOCcode/VOCinit.m line 81 修改类别的名字 CocoUtils.convertPascalGt( 'E:/data/jinju', '2007', 'val', 'E:/data/jinju/pascal_val2007.json') CocoUtils.convertPascalGt( 'E:/data/jinju', '2007', 'trainval', 'E:/data/jinju/pascal_train2007.json') @maggiequan

thank you, i will try it.

xingyizhou commented 5 years ago

Hi, You can use the VOC data in json format from the detectron repo https://github.com/facebookresearch/Detectron/tree/master/detectron/datasets/data.

maggiequan commented 5 years ago

Hi, You can use the VOC data in json format from the detectron repo https://github.com/facebookresearch/Detectron/tree/master/detectron/datasets/data.

i just wanna to use the pretrained model to test my own pictures to see the effect. heilow says: you can change the _load_data function in db/coco.py. You just need to save your image file names to the _image_ids variable and just set _detections to be an empty list. but i don not know how to change.

TEOEFW commented 5 years ago

do you know how to use the voc to train? i want to train my own data,but i only have the voc format