def main():
# load net
net_file = join(realpath(dirname(__file__)), 'SiamRPNBIG.model')
net = SiamRPNBIG()
net.load_state_dict(torch.load(net_file))
net.eval().cuda()
# warm up
for i in range(10):
net.temple(
torch.autograd.Variable(torch.FloatTensor(1, 3, 127, 127)).cuda())
net(torch.autograd.Variable(torch.FloatTensor(1, 3, 255, 255)).cuda())
# start to track
handle = vot.VOT("polygon")
Polygon = handle.region()
cx, cy, w, h = get_axis_aligned_bbox(Polygon)
image_file = handle.frame()
if not image_file:
sys.exit(0)
target_pos, target_sz = np.array([cx, cy]), np.array([w, h])
im = cv2.imread(image_file) # HxWxC
state = SiamRPN_init(im, target_pos, target_sz, net) # init tracker
while True:
image_file = handle.frame()
if not image_file:
break
im = cv2.imread(image_file) # HxWxC
state = SiamRPN_track(state, im) # track
res = cxy_wh_2_rect(state['target_pos'], state['target_sz'])
handle.report(Rectangle(res[0], res[1], res[2], res[3]))
def main():
vid_file = os.path.expanduser("~/Videos/VID_20190327_195111.mp4")
cap = cv2.VideoCapture(vid_file)
# load net
net = SiamRPNvot()
net.load_state_dict(
torch.load(join(realpath(dirname(__file__)), 'SiamRPNVOT.model')))
net.eval().cuda()
# # image and init box
# image_files = sorted(glob.glob('./bag/*.jpg'))
init_rbox = [
334.02, 128.36, 438.19, 188.78, 396.39, 260.83, 292.23, 200.41
]
[cx, cy, w, h] = get_axis_aligned_bbox(init_rbox)
# tracker init
target_pos, target_sz = np.array([cx, cy]), np.array([w, h])
ret, im = cap.read()
state = SiamRPN_init(im, target_pos, target_sz, net, use_gpu=True)
toc = 0
while (True):
# Capture frame-by-frame
ret, im = cap.read()
tic = cv2.getTickCount()
state = SiamRPN_track(state, im, use_gpu=True) # track
toc += cv2.getTickCount() - tic
res = cxy_wh_2_rect(state['target_pos'], state['target_sz'])
res = [int(l) for l in res]
cv2.rectangle(im, (res[0], res[1]), (res[0] + res[2], res[1] + res[3]),
(0, 255, 255), 3)
cv2.imshow('SiamRPN', im)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
# When everything done, release the capture
cap.release()
cv2.destroyAllWindows()
def OTB_run(gt_rects, frame_paths, tracker):
tic = time.time()
# tracking loop
res = []
for idx, img_path in enumerate(frame_paths):
print('Frame', idx)
img = np.array(Image.open(frame_paths[idx]).convert('RGB'))
if idx == 0:
if len(gt_rects[0]) == 8:
init_bbox = get_axis_aligned_bbox(np.array(gt_rects[0]))
else:
init_bbox = gt_rects[0]
pred_bbox = tracker.initialize(img, init_bbox)
else:
pred_bbox = tracker.track(img)
res.append(pred_bbox)
fps = len(res) / (time.time() - tic)
success_overlap = compute_success_overlap(gt_rects, res)
print('success overlap: %.4f, fps:%.2f' % (success_overlap.mean(), fps))
def SiamRPN_load(image, boxes, txt_path):
if not os.path.exists(txt_path):
os.makedirs(txt_path)
multiTracker = cv2.MultiTracker_create()
net = SiamRPNotb()
net.load_state_dict(
torch.load(join(realpath(dirname(__file__)), 'SiamRPNOTB.model')))
net.eval().cuda()
states, labels = [], []
for bbox in boxes:
#init_rbox = [bbox[0],bbox[1],bbox[0]+bbox[2],bbox[1],bbox[0]+bbox[2],bbox[1]+bbox[3],bbox[0],bbox[1]+bbox[3]]
init_rbox = [
bbox[0], bbox[1], bbox[2], bbox[1], bbox[2], bbox[3], bbox[0],
bbox[3]
]
[cx, cy, w, h] = get_axis_aligned_bbox(init_rbox)
# print(cx, cy, w, h,'-----',init_rbox)
target_pos, target_sz = np.array([cx, cy]), np.array([w, h])
states.append(SiamRPN_init(image, target_pos, target_sz, net))
labels.append(bbox[-1])
return states, labels
parser.add_argument('--model',
metavar='model',
default='SiamRPNPPRes50',
type=str,
help='which model to use.')
args = parser.parse_args()
# load net
net = eval(args.model)()
load_net('./{}.pth'.format(args.model), net)
net.eval().cuda()
# image and init box
image_files = sorted(glob.glob('./testData/*.jpg'))
init_rbox = [3641, 1778, 3810, 1778, 3810, 2313, 3641, 2313]
[cx, cy, w, h] = get_axis_aligned_bbox(init_rbox)
# tracker init
target_pos, target_sz = np.array([cx, cy]), np.array([w, h])
im = cv2.imread(image_files[0]) # HxWxC
state = SiamRPN_init(im, target_pos, target_sz, net, args.model)
# tracking and visualization
toc = 0
for f, image_file in enumerate(image_files):
im = cv2.imread(image_file)
# print(im.shape)
tic = cv2.getTickCount()
state = SiamRPN_track(state, im) # track
toc += cv2.getTickCount() - tic
res = cxy_wh_2_rect(state['target_pos'], state['target_sz'])
# load net
net_file = join(realpath(dirname(__file__)), 'SiamRPNBIG.model')
net = SiamRPNBIG()
net.load_state_dict(torch.load(net_file))
net.eval().cuda()
# warm up
for i in range(10):
net.temple(torch.autograd.Variable(torch.FloatTensor(1, 3, 127, 127)).cuda())
net(torch.autograd.Variable(torch.FloatTensor(1, 3, 255, 255)).cuda())
# start to track
handle = vot.VOT("polygon")
Polygon = handle.region()
cx, cy, w, h = get_axis_aligned_bbox(Polygon)
image_file = handle.frame()
if not image_file:
sys.exit(0)
target_pos, target_sz = np.array([cx, cy]), np.array([w, h])
im = cv2.imread(image_file) # HxWxC
state = SiamRPN_init(im, target_pos, target_sz, net) # init tracker
while True:
image_file = handle.frame()
if not image_file:
break
im = cv2.imread(image_file) # HxWxC
state = SiamRPN_track(state, im) # track
net(torch.autograd.Variable(torch.FloatTensor(1, 3, 255, 255)).cuda())
# image_files = sorted(glob.glob('./bag/*.jpg'))
# image_files = sorted(glob.glob('./helicopter/*.jpg'))
# image_files = sorted(glob.glob('./ant/*.jpg'))
# image_files = sorted(glob.glob('./basketball/*.jpg'))
# image_files = sorted(glob.glob('./crossing/*.jpg'))
image_files = sorted(glob.glob('./CarScale/*.jpg'))
# start to track 来给第一帧的图像初始化坐标用的，在第一帧以后的后续帧，用来更新坐标信息
# init_rbox = [334.02,128.36,438.19,188.78,396.39,260.83,292.23,200.41]
# init_rbox = [345.86,216.9,576.78,222.71,572.99,373.17,342.07,367.35]
# init_rbox = [137.21,458.36,156.83,460.78,148.35,529.41,128.72,526.99]
# init_rbox = [423,169,423,273,463,169,463,273]
# init_rbox = [201,153,201,203,223,153,223,203]
init_rbox = [6,166,6,189,48,171,48,192]
[cx, cy, w, h] = get_axis_aligned_bbox(init_rbox) # 第一帧起始box。将坐标数据转换成 RPN 的格式
if not image_files:
sys.exit(0)
target_pos, target_sz = np.array([cx, cy]), np.array([w, h]) # box中心坐标和box高和宽
im = cv2.imread(image_files[0]) # HxWxC
# SiamRPN_init 构造状态结构体并运行模板分支
state = SiamRPN_init(im, target_pos, target_sz, net) # init tracker
for f, image_file in enumerate(image_files):
if not image_file:
break
im = cv2.imread(image_file) # HxWxC
# 运行检测分支并更新状态变量
files = sorted(files)
images = [
dir_name + "/" + file_name for file_name in files
if (file_name[-3:] == "jpg")
]
gt_txt = dir_name + "/" + "groundtruth.txt"
# with open(gt_txt, 'r') as f:
# gt_box_str = f.readline()
# gt_box = map(float,list(gt_box_str.split(',')))
region_type = "polygon"
region = convert_region(parse_region(open(gt_txt, 'r').readline()),
region_type)
#Polygon = region
cx, cy, w, h = get_axis_aligned_bbox(region)
frame = 0
image_file = images[frame]
if not image_file:
sys.exit(0)
target_pos, target_sz = np.array([cx, cy]), np.array([w, h])
im = cv2.imread(image_file) # HxWxC
cv2.imshow("Tracking", im)
cv2.waitKey(1)
print("SiamRPN init")
state = SiamRPN_init(im, target_pos, target_sz, net) # init tracker
print("begin tracking")
while True:
frame = frame + 1
video_length = vot.get_frame_length(video_name)
#ground truth bounding box
gts = vot.get_gts(video_name)
frame_tags = vot.get_frame_tags(video_name)
video_frames = vot.get_frames(video_name)
flow_dir = os.path.join(flow_dirs, video_name + '.txt')
img_dir = os.path.join(vot_dir, video_name,'color')
confidence_dir = os.path.join('/home/jianingq/backward_flow_confidence_vot/',video_name)
#initialize network
# image and init box
init_rbox = gts[0]
if(len(init_rbox) == 4):
[cx, cy], [w, h] = rect_2_cxy_wh(init_rbox)
else:
[cx, cy, w, h] = get_axis_aligned_bbox(init_rbox)
# tracker init
target_pos, target_sz = np.array([cx, cy]), np.array([w, h])
im = video_frames[0]# HxWxC
state = SiamRPN_init(im, target_pos, target_sz, net)
detection_box = [int(cx-w/2),int(cy-h/2),int(cx+w/2),int(cy+h/2)]
for i in range(0,video_length - 1):
#track
im1 = np.copy(video_frames[i])
im2 = np.copy(video_frames[i + 1])
entropy_data = np.load(os.path.join(confidence_dir,format(i+1, '08')+'_entropy.npy'))
flow = np.load(os.path.join(confidence_dir,format(i+1, '08')+'_flow.npy'))
warped_im1 = flow_warp(im1, flow)
# load net
net_file = join(realpath(dirname(__file__)), 'SiamRPNBIG.model')
net = SiamRPNBIG()
net.load_state_dict(torch.load(net_file)) # 加载模型参数
net.eval().cuda() # 对模型进行验证，cuda(device=None) 将所有模型参数和缓冲区移动到GPU
# warm up
# 预热，运行模板和检测分支10次，先屏蔽掉
# for i in range(10):
# net.temple(torch.autograd.Variable(torch.FloatTensor(1, 3, 127, 127)).cuda()) # selonsy:cuda()表示使用GPU进行计算,FloatTensor(1, 3, 127, 127)：浮点型四维张量
# net(torch.autograd.Variable(torch.FloatTensor(1, 3, 255, 255)).cuda())
# start to track
handle = vot.VOT("polygon")
Polygon = handle.region() # region：将配置消息发送到客户端并接收初始化区域和第一个图像的路径。其返回值为初始化区域。
cx, cy, w, h = get_axis_aligned_bbox(Polygon) # get_axis_aligned_bbox：将坐标数据转换成 RPN 的格式
image_file = handle.frame() # frame 函数从客户端获取帧（图像路径）
if not image_file:
sys.exit(0)
target_pos, target_sz = np.array([cx, cy]), np.array([w, h])
im = cv2.imread(image_file) # HxWxC
state = SiamRPN_init(im, target_pos, target_sz, net) # init tracker,SiamRPN_init:构造状态结构体并运行模板分支
# 从第一帧开始跟踪，表示很奇怪，难道直接给定的不准确么？ # selonsy：改进点
while True: # 进入跟踪循环
image_file = handle.frame()
if not image_file:
break
im = cv2.imread(image_file) # HxWxC
state = SiamRPN_track(state, im) # track,SiamRPN_track:运行检测分支并更新状态变量
