-
Notifications
You must be signed in to change notification settings - Fork 164
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
the VP part problem #24
Comments
I am also interested in the VP part. However, I did not find it predicted in deploy.prototxt. I wish I was wrong here. |
@cuicuizhang1989 @Frandre You can simply add the VP branch to deploy.prototxt as follows. layer {
name: "vp-conv"
type: "Convolution"
bottom: "L6d"
top: "vp-conv"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
convolution_param {
num_output: 128
kernel_size: 1
weight_filler {
type: "gaussian"
std: 0.01
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "vp-tile"
type: "Tiling"
bottom: "vp-conv"
top: "vp-conv-tiled"
tiling_param {
tile_dim: 8
}
}
# Pixel level softmax prediction.
layer {
name: "vp-prob"
type: "Softmax"
bottom: "vp-conv-tiled"
top: "vp-prob"
} |
@cuicuizhang1989 @Frandre FYI, the python code to localize the VP is as below. vp_prob = net.blobs['vp-prob'].data[0]
[vp_y, vp_x] = vp_quadrant(vp_prob)
def vp_quadrant(prob):
prob[0,:,:] = scipy.ndimage.filters.gaussian_filter(prob[0,:,:], 5)
prob[1,:,:] = scipy.ndimage.filters.gaussian_filter(prob[1,:,:], 5)
prob[2,:,:] = scipy.ndimage.filters.gaussian_filter(prob[2,:,:], 5)
prob[3,:,:] = scipy.ndimage.filters.gaussian_filter(prob[3,:,:], 5)
prob[4,:,:] = scipy.ndimage.filters.gaussian_filter(prob[4,:,:], 5)
P_avg = (1 - np.mean(prob[0,:,:])) / 4
avg_map = np.ones(prob[0,:,:].shape) * P_avg
loc_map = np.power((avg_map-prob[1,:,:]), 2) + np.power((avg_map-prob[2,:,:]), 2) + np.power((avg_map-prob[3,:,:]), 2) + np.power((avg_map-prob[4,:,:]), 2)
[y, x] = np.unravel_index(loc_map.argmin(), loc_map.shape)
return [y, x] |
@SeokjuLee Hi, as you have provided the VP branch deploy.prototxt, can you also provide VP branch train.prototxt ? |
as the paper said. can the first task VPP be provide?
The text was updated successfully, but these errors were encountered: