Open bkvie opened 6 years ago
Hi, the gif was made by @felixlaumon (from https://github.com/felixlaumon/deform-conv), for more details, please consult him.
With libraries such as imageio, one can easily generates gif image from a list of numpy images:
import imageio
images = []
# generate your images here, using matplotlib etc., then convert to numpy arrays
...
imageio.mimsave('/path/to/movie.gif', images)
Refer to here: https://stackoverflow.com/questions/753190/programmatically-generate-video-or-animated-gif-in-python?answertab=votes#tab-top and also here: http://www.icare.univ-lille1.fr/wiki/index.php/How_to_convert_a_matplotlib_figure_to_a_numpy_array_or_a_PIL_image
Making the gif is easy, I was wondering about visualizing the networks field of view, ie the point cloud simulating the convolutions
You can read this : https://medium.com/@phelixlau/notes-on-deformable-convolutional-networks-baaabbc11cf3
the read dots are sampling location, you can just plot the offset
generated here: https://github.com/oeway/pytorch-deform-conv/blob/master/torch_deform_conv/layers.py#L39
plot it and make a gif when deforming the input image.
It would be a nice PR to have if you succeed.
hello,can you help me , I can't visualizing the receptive field
@kirkzZ Sorry that I haven't worked on this repo for a while, and some users mentioned there is some issue with the implementation here. So maybe you can try others' implementation instead?
I want to visualizing the receptive field as points(or visualizing the sample points), could you give me some ideas
Hi,
is there visualization for the method, as seen in the provided gif file? How would one ideally implement this feature?