Realtime_multi-person_pose_estimation: what is the meaning of np = 56 # num_parts
xiangweifeng
All 9 comments
@xiangweifeng Hi, I also wonder what is the meaning of np =56, And ,what does np=56 has to do with np_in_lmdb=17 , could you tell me ?
feitiandemiaomi
on 14 Sep 2017
The COCO datasets have 17 body parts, so np_in_lmdb=17.
In cpm_data_transformer from 182~184,
if(np == 56){
int COCO_to_ours_1[18] = {1,6, 7,9,11, 6,8,10, 13,15,17, 12,14,16, 3,2,5,4};
int COCO_to_ours_2[18] = {1,7, 7,9,11, 6,8,10, 13,15,17, 12,14,16, 3,2,5,4};
the 6-‘left shoulder’ and 7-‘right shoulder’ generate a new part. So there is 18 heatmaps.
In cpm_data_transformer from 1172~1173,
int mid_1[19] = {2, 9, 10, 2, 12, 13, 2, 3, 4, 3, 2, 6, 7, 6, 2, 1, 1, 15, 16};
int mid_2[19] = {9, 10, 11, 12, 13, 14, 3, 4, 5, 17, 6, 7, 8, 18, 1, 15, 16, 17, 18};
there are 19 correspondences,as well as 19*2=38 vecmaps.
Finally, 18+38=56.
xiangweifeng
on 14 Sep 2017
@xiangweifeng Thanks for your reply , It is very cool !
feitiandemiaomi
on 15 Sep 2017
@xiangweifeng Thank you for your great explanation. I found that you are using mid_1[12] and mid_2[12]. (Guess these are left shoulder and left ear.) I'm curious that why you generate PAF between two of them, could you tell me?
noirmist
on 15 Sep 2017
That'not my definition, it's authour's. I think it can improve the stability of the PAF. You can define any kind of correspondence.
xiangweifeng
on 16 Sep 2017
I see. thank you a lot!
2017-09-16 17:26 GMT+09:00 xiangweifeng notifications@github.com:
That'not my definition, it's authour's. I think it can improve the
stability of the PAF. You can define any kind of correpondence.—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
https://github.com/ZheC/Realtime_Multi-Person_Pose_Estimation/issues/93#issuecomment-329954560,
or mute the thread
https://github.com/notifications/unsubscribe-auth/AIQ71QFsb6rYkQz8EZm8b-bdIg6EpXzGks5si4YdgaJpZM4PSFtQ
.
noirmist
on 17 Sep 2017
What is mid_1 and mid_2 mean?
int mid_1[19] = {2, 9, 10, 2, 12, 13, 2, 3, 4, 3, 2, 6, 7, 6, 2, 1, 1, 15, 16};
int mid_2[19] = {9, 10, 11, 12, 13, 14, 3, 4, 5, 17, 6, 7, 8, 18, 1, 15, 16, 17, 18};
liaowang0125
on 20 Oct 2017
@liaowang0125 Have you figured it out? Someone can kindly explain that?
hellojialee
on 14 Jan 2018
@USTClj 19 connecting limbs
liaowang0125
on 2 Mar 2018
Related issues
RuWang15
·
12Comments
yq1011
·
21Comments
wuchichung
·
3Comments
chuanheliu
·
3Comments
gaopeng-eugene
·
20Comments
Most helpful comment
The COCO datasets have 17 body parts, so np_in_lmdb=17.
In cpm_data_transformer from 182~184,
if(np == 56){
int COCO_to_ours_1[18] = {1,6, 7,9,11, 6,8,10, 13,15,17, 12,14,16, 3,2,5,4};
int COCO_to_ours_2[18] = {1,7, 7,9,11, 6,8,10, 13,15,17, 12,14,16, 3,2,5,4};
the 6-‘left shoulder’ and 7-‘right shoulder’ generate a new part. So there is 18 heatmaps.
In cpm_data_transformer from 1172~1173,
int mid_1[19] = {2, 9, 10, 2, 12, 13, 2, 3, 4, 3, 2, 6, 7, 6, 2, 1, 1, 15, 16};
int mid_2[19] = {9, 10, 11, 12, 13, 14, 3, 4, 5, 17, 6, 7, 8, 18, 1, 15, 16, 17, 18};
there are 19 correspondences,as well as 19*2=38 vecmaps.
Finally, 18+38=56.