oujieww

euclidean loss caffe2 python

3

i can not find this function. so i use Sub() Abs() and DotProduct() to get loss for N-D diff=model.net.Sub([blob_out, 'label'],'diff') adiff=model.net.Abs([diff],'adiff') loss = model.net.DotProduct([adiff,adiff],'loss') but it do not work,Can anyone...

training loss is strange,for openpose body_25 model training

9

I use that model to train my dataset ,train.prototxt as : layer { name: "data" type: "CPMData" top: "data" top: "label" data_param { source: "/home/oujie/Desktop/Realtime_Multi-Person_Pose_Estimation/training/dataset/lmdb/" batch_size: 8 backend: LMDB }...

why don't you use hdf5 to save data and labels. Is it better to use Lmdb？why u save some other information such as dataset name into mate_data.joints information is enough...

i use the input heatmap for instance , crop from global heatmaps .resized to 1*K*36*56, when i vislized the output of the PRN,is not right

the speed of this paper, i have some questions. Please help me.

1

in your paper's abstract ,"the fastest real time system with ∼23 frames/sec". And we can find "Keypoint and person detections take 35 ms while PRN takes 2 ms per instance"...

mean iu always is 0.03469 not change loss is go down

5

i try to train segnet on voc ,i use voc-fcn setting and change lr to 2.5e-4 for bacth one,but the mean_iu is 0.03469 for every epoch, loss is not go...

training memory fluctuates

1

is there any one have this kand of problem,I try to train voc-fcn ,but while training go on, the memory is changing all the time,from 2G to 11G,and i run...

your deepercut is about mulit person ,but u give single-person model,too.I want to know can i use your single-person model to meet mulit-person task? And when training detection,Is datasets must...

some question about attention module

1

thanks for sharing the code. i have a question about the attention module in "transformer_models.py" in your code , assume that there are 10 tokens in setence, i think the...

I am looking for python eval code for rtpose, i find many issues on rtpose ,maybe 0.577 should be the result of rtpose of first 2644 images of val2014? and...