Running model on random image about ct-net HOT 2 CLOSED

Hi, our model depends on several pretrained model to extract the low-level information of the human, including human keypoints, densepose descriptor and the body parsing.

For the efficiency of testing and training, we prepare these inputs first, just as the one mapping of deepfashion image to pose file. But our model is able to run on any random image pairs. To achieve this, you may cascade our model with the pretrained model and finetune the whole framework.

from ct-net.

@yf1019 Thanks for replying. Makes complete sense.
It will be great if you can guide me how can I generate those for my own dataset, because looking at individual pretrained models, their output seems to differ in format compared to how it has been used in this model.

from ct-net.