def run_SiamFC(seq, rp, saveimage):
x = seq.init_rect[0]
y = seq.init_rect[1]
w = seq.init_rect[2]
h = seq.init_rect[3]
tic = time.clock()
# starting tracking
tracker = SiamFCTracker(config.model_path, config.gpu_id)
res = []
for idx, frame in enumerate(seq.s_frames):
frame = cv2.cvtColor(cv2.imread(frame), cv2.COLOR_BGR2RGB)
if idx == 0:
bbox = (x, y, w, h)
tracker.init(frame, bbox)
bbox = (bbox[0], bbox[1], bbox[0] + bbox[2], bbox[1] + bbox[3]
) # 1-idx
else:
bbox = tracker.update(frame)
res.append(
(bbox[0], bbox[1], bbox[2] - bbox[0], bbox[3] - bbox[1])) # 1-idx
duration = time.clock() - tic
result = {}
result['res'] = res
result['type'] = 'rect'
result['fps'] = round(seq.len / duration, 3)
return result
def main(video_dir, gpu_id, model_path):
# load videos
filenames = sorted(glob.glob(os.path.join(video_dir, "img/*.jpg")),
key=lambda x: int(os.path.basename(x).split('.')[0]))
frames = [
cv2.cvtColor(cv2.imread(filename), cv2.COLOR_BGR2RGB)
for filename in filenames
]
gt_bboxes = pd.read_csv(os.path.join(video_dir, "groundtruth_rect.txt"),
sep='\t|,| ',
header=None,
names=['xmin', 'ymin', 'width', 'height'],
engine='python')
title = video_dir.split('/')[-1]
# starting tracking
tracker = SiamFCTracker(model_path, gpu_id)
for idx, frame in enumerate(frames):
if idx == 0:
bbox = gt_bboxes.iloc[0].values
tracker.init(frame, bbox)
bbox = (bbox[0] - 1, bbox[1] - 1, bbox[0] + bbox[2] - 1,
bbox[1] + bbox[3] - 1)
else:
bbox = tracker.update(frame)
# bbox xmin ymin xmax ymax
frame = cv2.rectangle(frame, (int(bbox[0]), int(bbox[1])),
(int(bbox[2]), int(bbox[3])), (0, 255, 0), 2)
gt_bbox = gt_bboxes.iloc[idx].values
gt_bbox = (gt_bbox[0], gt_bbox[1], gt_bbox[0] + gt_bbox[2],
gt_bbox[1] + gt_bbox[3])
frame = cv2.rectangle(
frame,
(int(gt_bbox[0] - 1), int(gt_bbox[1] - 1)), # 0-index
(int(gt_bbox[2] - 1), int(gt_bbox[3] - 1)),
(255, 0, 0),
1)
if len(frame.shape) == 3:
frame = cv2.cvtColor(frame, cv2.COLOR_RGB2BGR)
frame = cv2.putText(frame, str(idx), (5, 20),
cv2.FONT_HERSHEY_COMPLEX_SMALL, 1, (0, 255, 0), 1)
cv2.imshow(title, frame)
cv2.waitKey(30)
def main(video_dir, gpu_id, model_path):
# load videos
filenames = sorted(glob.glob(os.path.join(video_dir, "img/*.jpg")),
key=lambda x: int(os.path.basename(x).split('.')[0]))
frames = [cv2.cvtColor(cv2.imread(filename), cv2.COLOR_BGR2RGB) for filename in filenames]
first_frame = frames[0]
#extract person detections and sort it from largest to smallest
person_detections = list(set(filter(lambda x:x[1]=='person', yolo(first_frame))))
person_detections.sort(key=lambda x: (x[0][2] * x[0][3]), reverse=True)
title = video_dir.split('/')[-1]
# starting tracking
trackers = []
bboxes = []
bboxes_colours = []
for i in range(len(person_detections)):
trackers.append(SiamFCTracker(model_path, gpu_id))
bboxes.append(person_detections[i][0])
bboxes_colours.append((randint(0, 255) % 255, randint(0, 255), randint(0, 255)))
for idx, frame in enumerate(frames):
if idx == 0:
for t in range(len(trackers)):
trackers[t].init(frame, person_detections[t][0])
bboxes[t] = (bboxes[t][0]-1, bboxes[t][1]-1,
bboxes[t][0]+ bboxes[t][2]-1, bboxes[t][1]+ bboxes[t][3]-1)
else:
for t in range(len(trackers)):
bboxes[t] = trackers[t].update(frame)
# bbox xmin ymin xmax ymax
for t in range(len(trackers)):
frame = cv2.rectangle(frame,
(int(bboxes[t][0]), int(bboxes[t][1])),
(int(bboxes[t][2]), int(bboxes[t][3])),
bboxes_colours[t],
2)
if len(frame.shape) == 3:
frame = cv2.cvtColor(frame, cv2.COLOR_RGB2BGR)
frame = cv2.putText(frame, str(idx), (5, 20), cv2.FONT_HERSHEY_COMPLEX_SMALL, 1, (0, 255, 0), 1)
cv2.imshow(title, frame)
cv2.waitKey(30)
from __future__ import absolute_import #??
import os
from got10k.datasets import *
from siamfc import SiamFCTracker
# import multiprocessing
# multiprocessing.set_start_method('spawn',True)
if __name__ == '__main__':
root_dir = os.path.abspath('data/GOT-10k') #获取当前工作目录
seqs = GOT10k(root_dir, subset='train', return_meta=True)
tracker = SiamFCTracker(net_path='./models/CIResNet22_PRETRAIN.model',
train=True) #优化器，GPU，损失函数，网络模型
tracker.train_over(seqs)
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='siamfcres22 tracking')
#.add_argument('--dataset',default='VOT2018', type=str,help='datasets')
#parser.add_argument('--save_path', default='./results', type=str,help='config file')
#parser.add_argument('--snapshot', default=snapshot, type=str,help='snapshot of models to eval')
parser.add_argument('--model_path',
default='./models/siamfcres22_50.pth',
type=str,
help='eval one special video')
# parser.add_argument('--video', default='', type=str, help='eval one special video')
#parser.add_argument('--vis', action='store_true',help='whether visualzie result')
args = parser.parse_args()
tracker = SiamFCTracker(net_path=args.model_path, train=False) #初始化一个追踪器
# root_dir = os.path.abspath('datasets/OTB')
# e = ExperimentOTB(root_dir, version=2013)
root_dir = os.path.abspath('datasets/OTB')
e = ExperimentOTB(root_dir, version=2015)
# root_dir = os.path.abspath('datasets/UAV123')
# e = ExperimentUAV123(root_dir, version='UAV123')
# root_dir = os.path.abspath('datasets/UAV123')
# e = ExperimentUAV123(root_dir, version='UAV20L')
# root_dir = os.path.abspath('datasets/DTB70')
# e = ExperimentDTB70(root_dir)
from __future__ import absolute_import #??
import os
import sys
sys.path.append(os.getcwd())
from got10k.datasets import *
from siamfc import SiamFCTracker
os.environ["CUDA_VISIBLE_DEVICES"] = "0,1" #多卡情况下默认多卡训练,如果想单卡训练,设置为"0"
if __name__ == '__main__':
root_dir = os.path.abspath('data/GOT-10k') #获取当前工作目录
seqs = GOT10k(root_dir, subset='train', return_meta=True)
# group=2 是分组卷积, group=1 是原始的alexnet
tracker = SiamFCTracker(model_path='./models/alexnet.pth')
tracker.train_over(seqs)
#parser.add_argument('--snapshot', default=snapshot, type=str,help='snapshot of models to eval')
parser.add_argument('--model_path',
default='./models/siamfc_41.pth',
type=str,
help='eval one special video')
# parser.add_argument('--video', default='', type=str, help='eval one special video')
#parser.add_argument('--vis', action='store_true',help='whether visualzie result')
args = parser.parse_args()
# OTB100 单机,单GPU,8b,siamfc_50.pth -prec_score:0.790 -succ_score:0.591 -succ_rate:0.735
# OTB100 单机,2-GPU,8b,siamfc_50.pth -prec_score:0.756 -succ_score:0.563 -succ_rate:0.699
# OTB100 单机,2-GPU,16b,siamfc_50.pth -prec_score:0.760 -succ_score:0.575 -succ_rate:0.721
# OTB100 单机,2-GPU,16b,siamfc_38.pth -prec_score:0.755 -succ_score:0.569 -succ_rate:0.711
tracker = SiamFCTracker(model_path=args.model_path) #初始化一个追踪器
# root_dir = os.path.abspath('datasets/OTB')
# e = ExperimentOTB(root_dir, version=2013)
root_dir = os.path.abspath('datasets/OTB')
e = ExperimentOTB(root_dir, version=2015)
# root_dir = os.path.abspath('datasets/UAV123')
# e = ExperimentUAV123(root_dir, version='UAV123')
# root_dir = os.path.abspath('datasets/UAV123')
# e = ExperimentUAV123(root_dir, version='UAV20L')
# root_dir = os.path.abspath('datasets/DTB70')
# e = ExperimentDTB70(root_dir)
import multiprocessing
multiprocessing.set_start_method('spawn', True)
gpu_id = 0
model_path = './models/siamfc_30.pth'
# model_gpu=nn.DataParallel(model,device_ids=[0,1])# 多GPU并行计算
# output=model_gpu(input)
if __name__ == '__main__':
tracker = SiamFCTracker(model_path, gpu_id) #初始化一个追踪器
# root_dir = os.path.abspath('datasets/OTB')
# e = ExperimentOTB(root_dir, version=2013)
# root_dir = os.path.abspath('datasets/OTB')
# e = ExperimentOTB(root_dir, version=2015)
root_dir = os.path.abspath('datasets/UAV123')
e = ExperimentUAV123(root_dir, version='UAV123')
# root_dir = os.path.abspath('datasets/UAV123')
# e = ExperimentUAV123(root_dir, version='UAV20L')
# root_dir = os.path.abspath('datasets/DTB70')
# e = ExperimentDTB70(root_dir)
from __future__ import absolute_import
import os
import glob
import numpy as np
from siamfc import SiamFCTracker
if __name__ == '__main__':
seq_dir = os.path.expanduser('~/data/OTB/Crossing')
img_files = sorted(glob.glob(seq_dir + 'img/*.jpg'))
anno = np.loadtxt(seq_dir + 'groundtruth_rect.txt')
net_path = './models/siamfcres22_50.pth'
tracker = SiamFCTracker(net_path=net_path)
tracker.track(img_files, anno[0], visualize=True)
def main():
# parse arguments
parser = argparse.ArgumentParser(description="Demo SiamFC")
parser.add_argument('-v',
'--video',
default='./video/Basketball',
type=str,
help="the path of video directory")
parser.add_argument('-g',
'--gpuid',
default=0,
type=int,
help="the id of GPU")
parser.add_argument('-m',
'--model',
default='./models/siamfc_pretrained.pth',
type=str,
help="the path of model")
args = parser.parse_args()
video_dir = args.video
gpu_id = args.gpuid
model_path = args.model
# load videos
filenames = sorted(glob.glob(os.path.join(video_dir, "img/*.jpg")),
key=lambda x: int(os.path.basename(x).split('.')[0]))
frames = [
cv2.cvtColor(cv2.imread(filename), cv2.COLOR_BGR2RGB)
for filename in filenames
]
gt_bboxes = pd.read_csv(os.path.join(video_dir, "groundtruth_rect.txt"),
sep='\t|,| ',
header=None,
names=['xmin', 'ymin', 'width', 'height'],
engine='python')
title = video_dir.split('/')[-1]
# starting tracking
tracker = SiamFCTracker(model_path, gpu_id)
for idx, frame in enumerate(frames):
if idx == 0:
bbox = gt_bboxes.iloc[0].values
tracker.init(frame, bbox)
bbox = (bbox[0] - 1, bbox[1] - 1, bbox[0] + bbox[2] - 1,
bbox[1] + bbox[3] - 1)
else:
bbox = tracker.update(frame)
# bbox xmin ymin xmax ymax
frame = cv2.rectangle(frame, (int(bbox[0]), int(bbox[1])),
(int(bbox[2]), int(bbox[3])), (0, 255, 0), 2)
gt_bbox = gt_bboxes.iloc[idx].values
gt_bbox = (gt_bbox[0], gt_bbox[1], gt_bbox[0] + gt_bbox[2],
gt_bbox[1] + gt_bbox[3])
frame = cv2.rectangle(
frame,
(int(gt_bbox[0] - 1), int(gt_bbox[1] - 1)), # 0-index
(int(gt_bbox[2] - 1), int(gt_bbox[3] - 1)),
(255, 0, 0),
1)
if len(frame.shape) == 3:
frame = cv2.cvtColor(frame, cv2.COLOR_RGB2BGR)
frame = cv2.putText(frame, str(idx), (5, 20),
cv2.FONT_HERSHEY_COMPLEX_SMALL, 1, (0, 255, 0), 1)
cv2.imshow(title, frame)
cv2.waitKey(30)
for i in args.instance_size:
cfg['instance_size'] = i
for l in args.scale_lr:
cfg['scale_lr'] = l
for k in args.scale_penalty:
cfg['scale_penalty'] = k
for w in args.window_influence:
cfg['window_influence'] = w
count += 1
tracker = SiamFCTracker(args.model_path, True, 0, cfg)
# root_dir = os.path.abspath('datasets/OTB')
# e = ExperimentOTB(root_dir, version=2013)
root_dir = os.path.abspath('datasets/OTB')
e = ExperimentOTB(root_dir, version=2015)
# root_dir = os.path.abspath('datasets/UAV123')
# e = ExperimentUAV123(root_dir, version='UAV123')
# root_dir = os.path.abspath('datasets/UAV123')
# e = ExperimentUAV123(root_dir, version='UAV20L')
# root_dir = os.path.abspath('datasets/DTB70')
# e = ExperimentDTB70(root_dir)
for i in args.instance_size:
cfg['instance_size'] = i
for l in args.scale_lr:
cfg['scale_lr'] = l
for k in args.scale_penalty:
cfg['scale_penalty'] = k
for w in args.window_influence:
cfg['window_influence'] = w
count += 1
tracker = SiamFCTracker(args.model_path, cfg)
# root_dir = os.path.abspath('datasets/OTB')
# e = ExperimentOTB(root_dir, version=2013)
root_dir = os.path.abspath('datasets/OTB')
e = ExperimentOTB(root_dir, version=2015)
# root_dir = os.path.abspath('datasets/UAV123')
# e = ExperimentUAV123(root_dir, version='UAV123')
# root_dir = os.path.abspath('datasets/UAV123')
# e = ExperimentUAV123(root_dir, version='UAV20L')
# root_dir = os.path.abspath('datasets/DTB70')
# e = ExperimentDTB70(root_dir)