search

mks0601 / PoseFix_RELEASE

Official TensorFlow implementation of "PoseFix: Model-agnostic General Human Pose Refinement Network", CVPR 2019
MIT License
329 stars 64 forks source link

demo for a single image #39

Open hosnasattar opened 4 years ago

hosnasattar commented 4 years ago

could you provide a demo running on a single image?

snitchjinx commented 4 years ago

Hello! I wonder if it's feasible to test the pre-trained COCO-based posefix on a name_of_input_pose.json outside the COCO dataset. Is the result in name_of_input_pose.json confined to image samples from COCO dataset? Thanks!

mks0601 commented 4 years ago

Sorry for late reply. @hosnasattar For the single image input, you only need to hack dataset.py. For example, you can edit data/COCO/dataset.py to load a single image and you can run the same command described in the README. Could you try this trick?

@snitchjinx Like the above answer, you can hack dataset.py. The pre-trained PoseFix model is not restricted to the images in the COCO dataset. You can try any images with the initial pose.

hosnasattar commented 4 years ago

@mks0601 I can give it a try but I think it would be great in general if you could provide a demo for a single image.

mks0601 commented 4 years ago

Yeah I agree but I'm busy doing other projects :( I'll try when I have time. Thanks!