def showNearestMHIs(testMoments, trainMoments, trainDirectoryNames, K):
distance = np.zeros((1, trainMoments.shape[0]))
nMHI = np.zeros((480, 640, K + 1))
for i in range(0, trainMoments.shape[0]):
distance[:, i] = np.sqrt(
np.sum(
np.divide(
np.power(
np.reshape(trainMoments[i, :],
(-1, 1)) - np.reshape(testMoments,
(-1, 1)), 2),
np.reshape(np.nanvar(trainMoments, axis=0), (-1, 1)))))
sorted_train_directory_names = np.reshape(
trainDirectoryNames[np.argsort(distance)], (trainDirectoryNames.shape))
for i in range(0, K + 1):
nMHI[:, :, i] = computeMHI(sorted_train_directory_names[i, :][0])
return nMHI
def showNearestMHI(testMoments, trainMoments, train_dir,K):
train_set_len = trainMoments.shape[0]
variance = np.nanvar(trainMoments, axis=0).reshape(-1, 1)
# variance=variance.reshape(20, 1)
distance = np.zeros((1, train_set_len))
testMoments = np.reshape(testMoments, (-1, 1))
# variance=variance.reshape(20, 1)
nearest_MHI=np.zeros((480,640,K+1))
for i in range(0, train_set_len):
distance[:,i] = normalized_euclidean(testMoments,trainMoments,variance,i)
sorted_index = np.argsort(distance)
#predictedAction = np.int(trainLabels[sorted_index][0])
#sorted directory
shape = train_dir.shape
sorted_directory=train_dir[sorted_index].reshape(shape)
#find k image
for i in range (0,K+1):
print("compute MHI for ", i ,"Nearest ")
nearest_MHI[:,:,i]= computeMHI(np.str(sorted_directory[i,:][0]))
return nearest_MHI
import matplotlib.pyplot as plt
import numpy as np
from computeMHI import computeMHI
from data_paths import seqpaths
def display_mhi(mhi, action):
plt.figure()
plt.imshow(mhi, cmap='gray')
plt.title(action)
plt.xticks([])
plt.yticks([])
plt.savefig(action + "_mhi.png")
plt.show()
if __name__ == '__main__':
# generate all MHIs
allMHIs_list = []
for path in seqpaths:
H = computeMHI(path)
allMHIs_list.append(H)
allMHIs = np.stack(allMHIs_list, axis=-1)
# dump MHIs
with open('allMHIs.npy', "wb") as f:
pickle.dump(allMHIs, f)
import cv2
import os
from predictAction import predictAction
from computeMHI import computeMHI
from huMoments import huMoments
basedir = './'
actions = ['botharms', 'crouch', 'leftarmup', 'punch', 'rightkick']
allMHIs_counter = 0
allMHIs = np.empty((480,640,20), dtype=np.float32)
allMoments = np.empty((20,7), dtype=np.float64)
for actionnum in range(len(actions)):
subdirname = basedir + actions[actionnum] + '/'
subdir = os.listdir(subdirname)
for seqnum in range(len(subdir)):
allMHIs[:,:,allMHIs_counter] = computeMHI(subdirname + subdir[seqnum])
allMoments[allMHIs_counter,:] = huMoments(allMHIs[:,:,allMHIs_counter])
allMHIs_counter += 1
allLabels = [i for i in xrange(1,6) for j in xrange(4)]
allLabels = np.array(allLabels)
#%%
actions = ['botharms', 'crouch', 'leftarmup', 'punch', 'rightkick']
#actions = ['leftarmup']
#trainMHI = np.empty((480,640,20), dtype=np.float32)
#trainMHI_counter = 0
np.save("huVectors", allMoments)
np.save("allMHIs", allMHIs)
allMoments = allMoments/np.linalg.norm(allMoments)
confusionMatrix = np.zeros((5,5))
for actionnum in range(len(actions)):
from huMoments import huMoments
basedir = './'
actions = ['botharms', 'crouch', 'leftarmup', 'punch', 'rightkick']
for actionnum in range(len(actions)):
subdirname = basedir + actions[actionnum] + '/'
subdir = os.listdir(subdirname)
allMHIs = np.empty((480,640,20), dtype=np.float32)
allMHIs_counter = 0
for seqnum in range(len(subdir)):
# cycle through all sequences for this action category
allMHIs_counter = allMHIs_counter + 1
allMHIs[:,:,allMHIs_counter] = computeMHI(subdirname + subdir[seqnum])
cv2.imshow(actions[actionnum]+"- %d"% seqnum, allMHIs[:,:,allMHIs_counter])
# cv2.imshow("mask", motion_mask)
if 0xFF & cv2.waitKey(1) == 27:
break
#cv2.destroyAllWindows()
#%%
allHuMoments = np.empty((20,7), dtype=np.float32)
for i in range(allMHIs.shape[2]):
allHuMoments[i,:] = huMoments(allMHIs[:,:,i])
#%%
"""
Created on Thu Nov 28 18:11:16 2019
@author: songl
"""
from computeMHI import computeMHI
import matplotlib.pyplot as plt
import numpy as np
import os
MHI = computeMHI('./PS5_Data/botharms/botharms-up-p1-1')
plt.title('Both Arms')
plt.imshow(MHI)
plt.show()
plt.imsave('botharms-up-p1-1_MHI.png', MHI)
np.save('botharms-up-p1-1_MHI.npy', MHI)
MHI = computeMHI('./PS5_Data/crouch/crouch-p1-1')
plt.title('Crouch')
plt.imshow(MHI)
plt.show()
plt.imsave('crouch-p1-1_MHI.png', MHI)
np.save('crouch-p1-1_MHI.npy', MHI)
MHI = computeMHI('./PS5_Data/leftarmup/leftarm-up-p1-1')
plt.title('Left Arm Up')
plt.imshow(MHI)
plt.show()
plt.imsave('leftarm-up-p1-1_MHI.png', MHI)
np.save('leftarm-up-p1-1_MHI.npy', MHI)
from huMoments import huMoments
basedir = './'
actions = ['botharms', 'crouch', 'leftarmup', 'punch', 'rightkick']
allMHIs = list()
for action in actions:
subdirname = basedir + action + '/'
subdirs = os.listdir(subdirname)
for seq in subdirs:
# cycle through all sequences for this action category
directory = subdirname + seq
mhi = computeMHI(directory)
allMHIs.append(mhi)
allMHIs = np.stack(allMHIs, axis=2)
np.save('allMHIs.npy', allMHIs)
print 'saved MHIs'
allHus = list()
allMHIs = np.load('allMHIs.npy')
_, _, numSeqs = np.shape(allMHIs)
for i in range(numSeqs):
mhi = allMHIs[:, :, i]
huVec = huMoments(mhi)
allHus.append(huVec)