def DeepKCF(seq, res_path, bSaveImage):
parameters = Params()
#img_files, pos, target_sz, ground_truth, video_path = load_video_info(video_path) # Load video info
img_files, pos, target_sz, video_path = load_video_gt(seq)
parameters.init_pos = np.floor(pos) + np.floor(
target_sz / 2) # Initial position
parameters.pos = parameters.init_pos # Current position
parameters.target_size = np.floor(target_sz) # Size of target
parameters.img_files = img_files # List of image files
parameters.video_path = video_path # Path to the sequence
num_frames = len(img_files)
results = np.zeros((num_frames, 4))
start = start_timer()
# For each frame
for frame in xrange(num_frames):
# Read the image
#im = cv2.imread(video_path + img_files[frame], 1)
im = scpmi.imread(video_path + img_files[frame])
# Initialize the tracker using the first frame
if frame == 0:
tracker1 = Tracker(im, parameters)
tracker1.train(im, True)
results[frame, :] = np.array((pos[0] + np.floor(target_sz[0] / 2),
pos[1] + np.floor(target_sz[1] / 2),
target_sz[0], target_sz[1]))
else:
results[frame, :], lost, xtf = tracker1.detect(
im) # Detect the target in the next frame
if not lost:
tracker1.train(im, False,
xtf) # Update the model with the new infomation
tracker1.close()
duration = end_timer(start, "to complete tracking")
fps = round(num_frames / duration, 2)
return results, fps
image = cv2.imread(imagefile, cv2.IMREAD_COLOR)
parameters = Params()
parameters.init_pos = np.floor([selection.x+selection.width/2, selection.y+selection.height/2]) # Initial position
parameters.pos = parameters.init_pos
parameters.target_size = np.floor([selection.width, selection.height])
tracker = Tracker(image, parameters)
tracker.train(image, True)
while True:
# *****************************************
# VOT: Call frame method to get path of the
# current image frame. If the result is
# null, the sequence is over.
# *****************************************
imagefile = handle.frame()
if not imagefile:
break
# Processing the current image
image = cv2.imread(imagefile, cv2.IMREAD_COLOR)
region, lost, xtf = tracker.detect(image)
if not lost:
tracker.train(image, False, xtf)
# *****************************************
# VOT: Report the position of the object
# every frame using report method.
# *****************************************
handle.report(selection)
def main():
# video_path = select_sequence()
video_path = '/home/tiago/data_seq/Car2'
parameters = Params()
img_files, pos, target_sz, ground_truth, video_path = load_video_info(
video_path) # Load video info
parameters.init_pos = np.floor(pos) + np.floor(
target_sz / 2) # Initial position
parameters.pos = parameters.init_pos # Current position
parameters.target_size = np.floor(target_sz) # Size of target
parameters.img_files = img_files # List of image files
parameters.video_path = video_path # Path to the sequence
num_frames = len(img_files)
results = np.zeros((num_frames, 4))
start = start_timer()
# For each frame
for frame in xrange(num_frames):
# Read the image
im = cv2.imread(video_path + img_files[frame], 1)
# Initialize the tracker using the first frame
if frame == 0:
tracker1 = Tracker(im, parameters)
tracker1.train(im, True)
results[frame, :] = np.array((pos[0] + np.floor(target_sz[0] / 2),
pos[1] + np.floor(target_sz[1] / 2),
target_sz[0], target_sz[1]))
else:
results[frame, :], lost, xtf = tracker1.detect(
im) # Detect the target in the next frame
if not lost:
tracker1.train(im, False,
xtf) # Update the model with the new infomation
if parameters.visualization:
# Draw a rectangle in the estimated location and show the result
cvrect = np.array((results[frame, 1] - results[frame, 3] / 2,
results[frame, 0] - results[frame, 2] / 2,
results[frame, 1] + results[frame, 3] / 2,
results[frame, 0] + results[frame, 2] / 2))
cv2.rectangle(im, (cvrect[0].astype(int), cvrect[1].astype(int)),
(cvrect[2].astype(int), cvrect[3].astype(int)),
(0, 255, 0), 2)
cv2.imshow('Window', im)
cv2.waitKey(1)
print(frame, end='\t')
duration = end_timer(start, "to complete tracking")
fps = round(num_frames / duration, 2)
print(fps, "FPS")
np.savetxt('results.txt', results, delimiter=',', fmt='%d')
if parameters.debug:
plt.figure()
plt.plot(tracker1.confidence)
plt.draw()
plt.figure()
plt.plot(tracker1.psr)
plt.draw()
plt.figure()
plt.plot(tracker1.high_freq_energy)
plt.show()
