how do I make my own data set to generate LMDB witch likes COCO for openpose ?

[ktr-hubrt]

i also want to address the problem,i found that the COCO lmdb has 6 parts:3d rgb, meta data,mask_all and mask_miss,i want to know how to change the form of data inputed to the net?

[feitiandemiaomi]

@Gzzgz @ktr-hubrt Hi, Do you solve it?

[oujieww]

I also have this problem,I want to know can i just make my dataset with 'input_image/3d_rgb' and '16_heatmaps as label' to setup my dataset.maybe this way will be different with the trainning of auther's.

[feitiandemiaomi]

@oujieww You are right , I try to train but failed such as : blob.cpp:32 Check failed :shape[i]>=0 (-13 vs 0)

[oujieww]

@feitiandemiaomi oho.yes,did u use lmbd?there too many other labels needed in author's model.I want to change some thing ,maybe to setup my dataset just with 'data_image' and 'label_16_heatmaps'.I want to have a try ,我感觉这样做应该需要把模型先修改下才可以

[Ai-is-light]

@Gzzgz @oujieww Did you get it?

[salvador-blanco]

I also want to use my own data set, I wonder if there is a tool to create a COCO like dataset/annotations? @Gzzgz @oujieww @Ai-is-light were you able to find a solution?