Khanh Ha

icon location Computer vision, Computer Graphics. Specialized in technologies for digital humans, from 3D avatar estimation to motion capture

Results 18 comments of Khanh Ha

caffe2 GPU support issue

after reading cmakelist.txt, I found out that Caffe is now a submodule, so we have to use the command " git submodule update --init --recursive" after pulling openpose.

input image normalization

Thanks for your answer. -> The output scale does not have to be the same as that of the input: --> Sorry for the misunderstanding. I mean the scale value...

input image normalization

Thank you very much

the reason behind the usage of render_onehot_heatmap for training target instead of render_gaussian_heatmap

Sorry for wasting your time. I should have read the paper more carefully! Best.

the pretrained weight of resnet is invalid

hello, sorry for the late response. I accidentally deleted the model from my drive. Working on retraining resnet23 and will upload the weight in two days

the pretrained weight of resnet is invalid

@gitunit : hello. Sorry. I am too busy at work. You might have to train resnet yourself. But the accuracy from Resnet is not much higher than VGG16.

Patchwise inference in unet_inference.py

Hi, this patch-based is just an experiment, not the approach that produces my published result.

About the Use of Silhouette

@YinghaoHuang91 : hi Ying. Thank you for your great work. Could you please share with me the code for the silhouette reprojection error through my email [email protected]? I find this...

stackwalker decoder fails to decode minidump file generated by crashpad_handler

hi @jameskr97, thanks for the quick response. here is the test data you need. - the .sym files are produced by the tool dump_sys - the .dmp file is produced...

stackwalker decoder fails to decode minidump file generated by crashpad_handler

@jameskr97 I am unable to see the output of crashserver-worker , I can only see the results in the browser, as shown in the screenshot I sent you