def main(opt, data_root='/data/MOT16/train', det_root=None, seqs=('MOT16-05',), exp_name='demo',
save_images=False, save_videos=False, show_image=True):
logger.setLevel(logging.INFO)
data_type = 'mot'
result_root = '/home/liujierui/proj/deep_sort_pytorch-master/demo/A-track/ensemble_MOT17_0.75'
# run tracking
accs = []
n_frame = 0
timer_avgs, timer_calls = [], []
for seq in seqs:
result_filename = os.path.join(result_root, '{}.txt'.format(seq))
# eval
logger.info('Evaluate seq: {}'.format(seq))
evaluator = Evaluator(data_root, seq, data_type)
accs.append(evaluator.eval_file(result_filename))
# get summary
metrics = mm.metrics.motchallenge_metrics
mh = mm.metrics.create()
summary = Evaluator.get_summary(accs, seqs, metrics)
strsummary = mm.io.render_summary(
summary,
formatters=mh.formatters,
namemap=mm.io.motchallenge_metric_names
)
print(strsummary)
def main(opt, data_root='/data/MOT16/train', det_root=None, seqs=('MOT16-05',), exp_name='demo',
save_images=False, save_videos=False, show_image=True):
logger.setLevel(logging.INFO)
result_root = os.path.join(data_root, '..', 'results', exp_name)
mkdir_if_missing(result_root)
data_type = 'mot'
print(opt)
# run tracking
accs = []
n_frame = 0
timer_avgs, timer_calls = [], []
for seq in seqs:
output_dir = os.path.join(data_root, '..', 'outputs', exp_name, seq) if save_images or save_videos else None
logger.info('start seq: {}'.format(seq))
# print(osp.join(data_root, seq, 'img1'))
dataloader = datasets.LoadImages(osp.join(data_root, seq), opt.img_size)
result_filename = os.path.join(result_root, '{}.txt'.format(seq))
meta_info = None #open(os.path.join(data_root, seq, 'seqinfo.ini')).read()
frame_rate = 15 #int(meta_info[meta_info.find('frameRate') + 10:meta_info.find('\nseqLength')])
nf, ta, tc = eval_seq(opt, dataloader, data_type, result_filename,
save_dir=output_dir, show_image=show_image, frame_rate=frame_rate)
n_frame += nf
timer_avgs.append(ta)
timer_calls.append(tc)
# eval
logger.info('Evaluate seq: {}'.format(seq))
evaluator = Evaluator(data_root, seq, data_type)
accs.append(evaluator.eval_file(result_filename))
if save_videos:
output_video_path = osp.join(output_dir, '{}.mp4'.format(seq))
cmd_str = 'ffmpeg -f image2 -i {}/%05d.jpg -c:v copy {}'.format(output_dir, output_video_path)
os.system(cmd_str)
timer_avgs = np.asarray(timer_avgs)
timer_calls = np.asarray(timer_calls)
all_time = np.dot(timer_avgs, timer_calls)
avg_time = all_time / np.sum(timer_calls)
logger.info('Time elapsed: {:.2f} seconds, FPS: {:.2f}'.format(all_time, 1.0 / avg_time))
# get summary
metrics = mm.metrics.motchallenge_metrics
mh = mm.metrics.create()
summary = Evaluator.get_summary(accs, seqs, metrics)
strsummary = mm.io.render_summary(
summary,
formatters=mh.formatters,
namemap=mm.io.motchallenge_metric_names
)
print(strsummary)
Evaluator.save_summary(summary, os.path.join(result_root, 'summary_{}.xlsx'.format(exp_name)))
def main(data_root='', seqs=('', ), args=""):
logger = get_logger()
logger.setLevel(logging.INFO)
data_type = 'mot'
result_root = os.path.join(Path(data_root), "mot_results")
#print("RESULT ROOT:",result_root)
mkdir_if_missing(result_root)
cfg = get_config()
#cfg.merge_from_file(args.config_detection)
cfg.merge_from_file(args.config_deepsort)
fcos = FCOSEvaluator("weights/FCOS_imprv_R_50_FPN_1x.pth",
config="./configs/fcos/fcos_imprv_R_50_FPN_1x.yaml")
#print("fcos.config",fcos.config)
fcos.prepare_detector() #build_detector(cfg, use_cuda=use_cuda)
# run tracking
accs = []
for seq in seqs:
logger.info('start seq: {}'.format(seq))
result_filename = "./tracking_output/results.txt" #os.path.join(result_root, '{}.txt'.format(seq))
video_path = data_root + "/" + seq + "/" + seq + ".mp4"
print(video_path)
with VideoTracker(cfg, args, video_path, fcos=fcos) as vdo_trk:
vdo_trk.run()
# eval
logger.info('Evaluate seq: {}'.format(seq))
evaluator = Evaluator(data_root, seq, data_type)
accs.append(evaluator.eval_file(result_filename))
# get summary
metrics = mm.metrics.motchallenge_metrics
mh = mm.metrics.create()
summary = Evaluator.get_summary(accs, seqs, metrics)
strsummary = mm.io.render_summary(summary,
formatters=mh.formatters,
namemap=mm.io.motchallenge_metric_names)
print(strsummary)
Evaluator.save_summary(summary,
os.path.join(result_root, 'summary_global.xlsx'))
def measure2files(results_root, data_root, seqs):
"""
@param results_root: mot轨迹预测结果文件的根目录，文件中格式 <frame>, <id>, <bb_left>, <bb_top>, <bb_width>, <bb_height>, <conf>, <x>, <y>, <z>
@param data_root: gt文件的路径，不包含后三级，因为在Evaluator初始化函数中已经写了全路径的拼接方式
@param seqs: gt路径的倒数第三级路径
@return: 存储seqs中，每一个路径下track结果的评价指标，以及全部文件夹汇总后的指标
"""
data_type = 'mot'
result_root = "/home/shuai.li/code/FairMOT/MOT15/images/results/temp/"
exp_name = "test_evalMot15"
accs = []
# eval
for seq in tqdm(seqs):
result_filename = osp.join(results_root, seq) + '.txt'
logger.info('Evaluate seq: {}'.format(seq))
evaluator = Evaluator(
data_root, seq,
data_type) # 在初始化中，根据data_root, seq自动加载ground truth数据
accs.append(evaluator.eval_file(
result_filename)) # 在读取存储的检测结果，并进行计算，一帧对应一个acc对象
# if save_videos:
# output_video_path = osp.join(output_dir, '{}.mp4'.format(seq))
# cmd_str = 'ffmpeg -f image2 -i {}/%05d.jpg -c:v copy {}'.format(output_dir, output_video_path)
# os.system(cmd_str)
# get summary
metrics = mm.metrics.motchallenge_metrics # 18个评价指标
mh = mm.metrics.create() # 创建指标计算工厂，后续传入相关数据即可给出指标
summary = Evaluator.get_summary(
accs, seqs, metrics) # 计算MOT challenge中的指标，参数：、eval的帧序列名称、指标序列
strsummary = mm.io.render_summary( # 将eval指标进行字符串格式化，用于在console上面的显示
summary,
formatters=mh.formatters,
namemap=mm.io.motchallenge_metric_names)
print(strsummary) # 显示在命令行中
Evaluator.save_summary(
summary, os.path.join(result_root, 'summary_{}.xlsx'.format(exp_name)))
def main(opt,
data_root='/data/MOT16/train',
det_root=None,
seqs=('MOT16-05', ),
exp_name='demo',
save_images=False,
save_videos=False,
show_image=True):
logger.setLevel(logging.INFO)
result_root = os.path.join(data_root, '..', 'results', exp_name)
mkdir_if_missing(result_root)
data_type = 'mot'
# run tracking
accs = []
n_frame = 0
timer_avgs, timer_calls = [], []
#scorre tutti gli elementi della sequenza di frame previsti dal MOT passato come parametro
for seq in seqs:
#stabilisce la cartella in cui dovrà salvare gli output elaborati
output_dir = os.path.join(data_root, '..', 'outputs', exp_name,
seq) if save_images or save_videos else None
#messaggio che viene restituito all'utente per aggiornarlo sull'andamento dell'esecuzione
logger.info('start seq: {}'.format(seq))
#carica l'immagine contenuta all'interno del dataset passato come parametro da seq
dataloader = datasets.LoadImages(osp.join(data_root, seq, 'img1'),
opt.img_size)
#Il file testuale con il risultato dell'elaborazione è salvato in result_filename
result_filename = os.path.join(result_root, '{}.txt'.format(seq))
#legge un file per seguire le indicazioni
meta_info = open(os.path.join(data_root, seq, 'seqinfo.ini')).read()
#qui avviene il calcolo del frame rate, per comprendere quanti frame fa nell'unità di tempo
frame_rate = int(meta_info[meta_info.find('frameRate') +
10:meta_info.find('\nseqLength')])
#esecuzione di eval_seq
nf, ta, tc = eval_seq(opt,
dataloader,
data_type,
result_filename,
save_dir=output_dir,
show_image=show_image,
frame_rate=frame_rate)
n_frame += nf
timer_avgs.append(ta)
timer_calls.append(tc)
# eval
logger.info('Evaluate seq: {}'.format(seq))
evaluator = Evaluator(data_root, seq, data_type)
accs.append(evaluator.eval_file(result_filename))
if save_videos:
output_video_path = osp.join(output_dir, '{}.mp4'.format(seq))
cmd_str = 'ffmpeg -f image2 -i {}/%05d.jpg -c:v copy {}'.format(
output_dir, output_video_path)
os.system(cmd_str)
timer_avgs = np.asarray(timer_avgs)
timer_calls = np.asarray(timer_calls)
all_time = np.dot(timer_avgs, timer_calls)
avg_time = all_time / np.sum(timer_calls)
logger.info('Time elapsed: {:.2f} seconds, FPS: {:.2f}'.format(
all_time, 1.0 / avg_time))
# get summary
metrics = mm.metrics.motchallenge_metrics
mh = mm.metrics.create()
summary = Evaluator.get_summary(accs, seqs, metrics)
strsummary = mm.io.render_summary(summary,
formatters=mh.formatters,
namemap=mm.io.motchallenge_metric_names)
print(strsummary)
Evaluator.save_summary(
summary, os.path.join(result_root, 'summary_{}.xlsx'.format(exp_name)))
def visualize_fp_fns_seq(result_dir,
dataroot,
seqs,
return_fp=True,
return_fn=True):
from tracking_utils.evaluation import Evaluator
accs = []
fp_info = []
tp_info = []
for seq in seqs:
img_dir = osp.join(dataroot, seq, 'img1')
output_dir = osp.join(f"{result_dir.replace('results', 'outputs')}",
'analyze', seq)
os.makedirs(output_dir, exist_ok=True)
evaluator = Evaluator(data_root=dataroot,
seq_name=seq,
data_type='mot',
return_fp=return_fp,
return_fn=return_fn)
if return_fp and return_fn:
acc, fns, fps, matched, fp_confs = evaluator.eval_file(
osp.join(result_dir, f"{seq}.txt"))
accs.append(acc)
# draw fp and fn
for frame_id in fps.keys():
fp = fps[frame_id]
fp_conf = fp_confs[frame_id]
fp[:, 2:] += fp[:, :2]
fn = fns[frame_id]
fn[:, 2:] += fn[:, :2]
match = matched[frame_id]
match['gt'][:, 2:] += match['gt'][:, :2]
match['trk'][:, 2:] += match['trk'][:, :2]
im = cv2.imread(osp.join(img_dir, f"{frame_id:06d}.jpg"))
for box, conf in zip(fp, fp_conf):
im = cv2.rectangle(im, tuple(box[0:2].astype(int)),
tuple(box[2:4].astype(int)),
(0, 0, 255), 2)
im = cv2.putText(im,
f"c:{conf:.3f}",
(int(box[0]), int(box[1] + 30)),
cv2.FONT_HERSHEY_PLAIN,
0.75, (0, 255, 255),
thickness=1)
height = box[3] - box[1]
im = cv2.putText(im,
f"h:{height:.3f}",
(int(box[0]), int(box[1] + 60)),
cv2.FONT_HERSHEY_PLAIN,
0.75, (0, 255, 255),
thickness=1)
fp_info.append([height, conf])
for box in fn:
im = cv2.rectangle(im, tuple(box[0:2].astype(int)),
tuple(box[2:4].astype(int)), (0, 0, 0),
2)
for gt, trk, conf in zip(match['gt'], match['trk'],
match['trk_confs']):
im = cv2.rectangle(im, tuple(gt[0:2].astype(int)),
tuple(gt[2:4].astype(int)), (0, 255, 0),
2)
im = cv2.rectangle(im, tuple(trk[0:2].astype(int)),
tuple(trk[2:4].astype(int)),
(255, 255, 0), 2)
im = cv2.putText(im,
f"c:{conf:.3f}",
(int(trk[0]), int(trk[1] + 30)),
cv2.FONT_HERSHEY_PLAIN,
0.75, (0, 255, 255),
thickness=1)
height = trk[3] - trk[1]
im = cv2.putText(im,
f"h:{height:.3f}",
(int(trk[0]), int(trk[1] + 60)),
cv2.FONT_HERSHEY_PLAIN,
0.75, (0, 255, 255),
thickness=1)
tp_info.append([height, conf])
cv2.imwrite(osp.join(output_dir, f"{frame_id:06d}.jpg"), im)
else:
accs.append(evaluator.eval_file(osp.join(result_dir,
f"{seq}.txt")))
metrics = mm.metrics.motchallenge_metrics
mh = mm.metrics.create()
summary = Evaluator.get_summary(accs, seqs, metrics)
strsummary = mm.io.render_summary(summary,
formatters=mh.formatters,
namemap=mm.io.motchallenge_metric_names)
print(strsummary)
fp_info = np.array(fp_info)
tp_info = np.array(tp_info)
import ipdb
ipdb.set_trace()
def main(opt,
data_root='/data/MOT16/train',
det_root=None,
seqs=('MOT16-05', ),
exp_name='demo',
save_images=False,
save_videos=False,
show_image=True):
logger.setLevel(logging.INFO)
# result_root = os.path.join(data_root, '..', 'results', exp_name)
result_root = os.path.join('/home/liujierui/proj/FairMOT-master/outputs',
exp_name)
mkdir_if_missing(result_root)
data_type = 'mot'
# run tracking
accs = []
n_frame = 0
timer_avgs, timer_calls = [], []
for seq in seqs:
output_dir = os.path.join(
'/home/liujierui/proj/FairMOT-master/outputs', exp_name,
seq) if save_images or save_videos else None
video_dir = os.path.join(
'/home/liujierui/proj/FairMOT-master/outputs',
exp_name) if save_images or save_videos else None
mkdir_if_missing(output_dir)
mkdir_if_missing(video_dir)
logger.info('start seq: {}'.format(seq))
dataloader = datasets.LoadImages(osp.join(data_root, seq, 'img1'),
opt.img_size)
result_filename = os.path.join(result_root, '{}.txt'.format(seq))
# meta_info = open(os.path.join(data_root, seq, 'seqinfo.ini')).read()
# frame_rate = int(meta_info[meta_info.find('frameRate') + 10:meta_info.find('\nseqLength')])
frame_rate = 15
nf, ta, tc = eval_seq(opt,
dataloader,
data_type,
result_filename,
save_dir=output_dir,
show_image=show_image,
frame_rate=frame_rate)
n_frame += nf
timer_avgs.append(ta)
timer_calls.append(tc)
# eval
logger.info('Evaluate seq: {}'.format(seq))
evaluator = Evaluator(data_root, seq, data_type)
accs.append(evaluator.eval_file(result_filename))
if save_videos:
# output_video_path = osp.join(video_dir, '{}.mp4'.format(seq))
# cmd_str = 'ffmpeg -f image2 -i {}/%05d.jpg -c:v copy {}'.format(output_dir, output_video_path)
# # cmd_str = 'ffmpeg -f image2 -i {}/%d.jpg -c:v copy {}'.format(output_dir, output_video_path)
# os.system(cmd_str)
path = output_dir
# path=join(self.video_path,str(id_video),'img1')
filelist_0 = os.listdir(path) #获取该目录下的所有文件名
filelist = [
'{:06d}'.format(i) for i in range(1,
len(filelist_0) + 1)
]
avi_name = join(video_dir, seq + ".avi") #导出路径
item = path + '/000001.jpg'
ori_im = cv2.imread(item)
video = cv2.VideoWriter(avi_name,
cv2.VideoWriter_fourcc('M', 'J', 'P', 'G'),
15, (ori_im.shape[1], ori_im.shape[0]))
for id_img in filelist:
item = path + '/' + id_img + '.jpg'
ori_im = cv2.imread(item)
video.write(ori_im)
video.release()
timer_avgs = np.asarray(timer_avgs)
timer_calls = np.asarray(timer_calls)
all_time = np.dot(timer_avgs, timer_calls)
avg_time = all_time / np.sum(timer_calls)
logger.info('Time elapsed: {:.2f} seconds, FPS: {:.2f}'.format(
all_time, 1.0 / avg_time))
# get summary
metrics = mm.metrics.motchallenge_metrics
mh = mm.metrics.create()
summary = Evaluator.get_summary(accs, seqs, metrics)
strsummary = mm.io.render_summary(summary,
formatters=mh.formatters,
namemap=mm.io.motchallenge_metric_names)
print(strsummary)
Evaluator.save_summary(
summary, os.path.join(result_root, 'summary_{}.xlsx'.format(exp_name)))
def main(opt,
data_root='/data/MOT16/train',
det_root=None,
seqs=('MOT16-05', ),
exp_name='demo',
save_images=False,
save_videos=False,
show_image=True):
logger.setLevel(logging.INFO)
result_root = os.path.join(opt.output_root, opt.exp_id)
mkdir_if_missing(result_root)
data_type = 'mot'
experiment = Experiment(api_key="SK59eWBf9ldDhEMbsQx7IW9G6",
project_name="fairmot",
workspace="noudvdgevel",
auto_param_logging=False,
auto_metric_logging=False,
auto_output_logging=False) #Comet experiment.
hyper_params = {
"conf_thres": opt.conf_thres,
"model": opt.load_model.split('/')[-1],
"data": opt.data_cfg,
"re_id_dim": opt.reid_dim,
"architecture": opt.arch
}
experiment.log_parameters(hyper_params)
experiment.set_name(opt.exp_id)
# run tracking
accs = []
n_frame = 0
timer_avgs, timer_calls = [], []
for seq in seqs:
output_dir = os.path.join(result_root, 'outputs',
seq) if save_images or save_videos else None
logger.info('start seq: {}'.format(seq))
meta_info = open(os.path.join(data_root, seq, 'seqinfo.ini')).read()
imDir = str(meta_info[meta_info.find('imDir') +
6:meta_info.find('\nframeRate')])
dataloader = datasets.LoadImages(osp.join(data_root, seq, imDir),
opt.img_size)
result_filename = os.path.join(result_root, '{}.txt'.format(seq))
frame_rate = int(meta_info[meta_info.find('frameRate') +
10:meta_info.find('\nseqLength')])
nf, ta, tc = eval_seq(opt,
dataloader,
data_type,
result_filename,
save_dir=output_dir,
show_image=show_image,
frame_rate=frame_rate)
n_frame += nf
timer_avgs.append(ta)
timer_calls.append(tc)
# eval
logger.info('Evaluate seq: {}'.format(seq))
evaluator = Evaluator(data_root, seq, data_type)
accs.append(evaluator.eval_file(result_filename))
if save_videos:
output_video_path = osp.join(output_dir, '{}.mp4'.format(seq))
cmd_str = 'ffmpeg -f image2 -i {}/%05d.jpg -c:v copy {}'.format(
output_dir, output_video_path)
os.system(cmd_str)
experiment.log_asset(output_video_path,
file_name="tracking_results.mp4",
copy_to_tmp=False)
timer_avgs = np.asarray(timer_avgs)
timer_calls = np.asarray(timer_calls)
all_time = np.dot(timer_avgs, timer_calls)
avg_time = all_time / np.sum(timer_calls)
logger.info('Time elapsed: {:.2f} seconds, FPS: {:.2f}'.format(
all_time, 1.0 / avg_time))
# get summary
metrics = mm.metrics.motchallenge_metrics
mh = mm.metrics.create()
summary = Evaluator.get_summary(accs, seqs, metrics)
strsummary = mm.io.render_summary(summary,
formatters=mh.formatters,
namemap=mm.io.motchallenge_metric_names)
summary_items = summary.to_numpy()
for c, seq in enumerate(seqs):
for cc, metric in enumerate(summary):
experiment.log_metric(metric + "_" + seq, summary_items[c][cc])
print(strsummary)
Evaluator.save_summary(
summary, os.path.join(result_root, 'summary_{}.xlsx'.format(exp_name)))
experiment.log_asset(os.path.join(result_root,
'summary_{}.xlsx'.format(exp_name)),
file_name="tracking_results.xlsx",
copy_to_tmp=False)
def main(
opt,
data_root="/data/MOT16/train",
det_root=None,
seqs=("MOT16-05", ),
exp_name="demo",
save_images=False,
save_videos=False,
show_image=True,
):
logger.setLevel(logging.INFO)
result_root = os.path.join(data_root, "..", "results", exp_name)
mkdir_if_missing(result_root)
data_type = "mot"
# run tracking
accs = []
n_frame = 0
timer_avgs, timer_calls = [], []
for seq in seqs:
output_dir = (os.path.join(data_root, "..", "outputs", exp_name, seq)
if save_images or save_videos else None)
logger.info("start seq: {}".format(seq))
dataloader = datasets.LoadImages(osp.join(data_root, seq, "img1"),
opt.img_size)
result_filename = os.path.join(result_root, "{}.txt".format(seq))
meta_info = open(os.path.join(data_root, seq, "seqinfo.ini")).read()
frame_rate = int(meta_info[meta_info.find("frameRate") +
10:meta_info.find("\nseqLength")])
nf, ta, tc = eval_seq(
opt,
dataloader,
data_type,
result_filename,
save_dir=output_dir,
show_image=show_image,
frame_rate=frame_rate,
)
n_frame += nf
timer_avgs.append(ta)
timer_calls.append(tc)
# eval
logger.info("Evaluate seq: {}".format(seq))
evaluator = Evaluator(data_root, seq, data_type)
accs.append(evaluator.eval_file(result_filename))
if save_videos:
output_video_path = osp.join(output_dir, "{}.mp4".format(seq))
cmd_str = "ffmpeg -f image2 -i {}/%05d.jpg -c:v copy {}".format(
output_dir, output_video_path)
os.system(cmd_str)
timer_avgs = np.asarray(timer_avgs)
timer_calls = np.asarray(timer_calls)
all_time = np.dot(timer_avgs, timer_calls)
avg_time = all_time / np.sum(timer_calls)
logger.info("Time elapsed: {:.2f} seconds, FPS: {:.2f}".format(
all_time, 1.0 / avg_time))
# get summary
metrics = mm.metrics.motchallenge_metrics
mh = mm.metrics.create()
summary = Evaluator.get_summary(accs, seqs, metrics)
strsummary = mm.io.render_summary(summary,
formatters=mh.formatters,
namemap=mm.io.motchallenge_metric_names)
print(strsummary)
Evaluator.save_summary(
summary, os.path.join(result_root, "summary_{}.xlsx".format(exp_name)))