regression box bug？

[LittlePeng]

            bbox.height = bounding_box_[j].height + regression_box_temp_[4*j+2] * bounding_box_[j].height;
            bbox.width = bounding_box_[j].width + regression_box_temp_[4*j+3] * bounding_box_[j].width;

看mathlab 代码，regression box 对应的应该是 x1,y1; x2,y2 两个坐标，而不是height，width 造成最终结果的bbox位置会有点歪，改正后效果更好些

[GarrickLin]

就觉得怎么抖得那么厉害，改完以后好很多

[wuzhiyang2016]

//regression box : y x height width regression_box.push_back(cv::Rect(regression_box_temp_[i + count] * width, regression_box_temp_[i] * width, regression_box_temp_[i + count3] * width, regression_box_temp_[i + count2] * width));

hi, could you tell me , in MTCNN.cpp, why save as [y,x,h,w] ? instead of [x,y,w,h]

[wuzhiyang2016]

At the train time, the label regress as [x,y,w,h],,,at the test time ,,,it must the order [x,y ,w,h],,,but in the code, why save as [y,x,h,w] ??? it is so confused!!!

[wuzhiyang2016]

@foreverYoungGitHub @LittlePeng @GarrickLin

[GarrickLin]

Data layout in caffe follows NCWH, but opencv is CHW. Before forwarding there is a transpose, so what you get then is all transposed.

[wuzhiyang2016]

Very thanks,i get it! @GarrickLin

[edric1261234]

@GarrickLin 能把改过的代码贴出来吗？