fmahoudeau / shelfnet-human-pose-estimation Goto Github PK

Hi, and thank you for making this code available.
Is there an example to run inference on a single image using cv?

Would this be a simple task to implement?

Thanks!

Hi,

I am unable to find/download the model.

Downloading /home/fanos/Documents/ModelZoo/official\resnet50-853f2fb0.zip from https://hangzh.s3.amazonaws.com/encoding/models/resnet50-853f2fb0.zip.

I get the following error

RuntimeError: Failed downloading url https://hangzh.s3.amazonaws.com/encoding/models/resnet50-853f2fb0.zip

Kindly somebody help me.

It's a great job! thank you so much !
but when i test the model on coco data ,the map is 0.655,which is much lower than 0.72. use the model you provided.

Hello!
why the train acc is always is 0.00 when i train the shelfnet50 ?can you tell me why?

Hi @fmahoudeau! Thank you for sharing ws such interesting solution. I wonder how dow you estimate its stability in near profile poses? Does it supports hard self occlusion situations without omitting predict of some keypoints like openpose did?
Another issue of ShelfNet backbone - how do you estimate its ability for DensePose task in scope of low inference time?