FairMOT训练kittitracking数据集的汽车类（参考FairVehicle）
⼯作情况：
kitti数据集的标签（转换成FairMOT类似的gt）
以下是处理图⽚和gt⽂本的程序操作，仅做个⼈记录
1.将原先kitti tracking数据集的标签修改去掉type⽤⽂本来表⽰改成数字代表同时将空格改成逗号
效果图
kitti tracking 标签含义：
The label files contain the following information, which can be read and written using the matlab tools (readLabels.m) provided within this devkit. All values (numerical or strings) are parated via spaces, each row corresponds to one object. The 17 columns reprent:
#Values Name Description
----------------------------------------------------------------------------
1 frame Frame within the quence where the object appearers 1 track id Unique tracking id of this object within this quence
1 type Describes the type of object:'Car','Van','Truck',
'Pedestrian','Person_sitting','Cyclist','Tram',
'Misc'or'DontCare'
1 truncated Integer (0,1,2) indicating the level of truncation.
Note that this is in contrast to the object detection
benchmark where truncation is a float in [0,1].
1 occluded Integer (0,1,2,3) indicating occlusion state:
0= fully visible,1= partly occluded
2= largely occluded,3= unknown
1 alpha Obrvation angle of object, ranging [-pi..pi]
4 bbox 2D bounding box of object in the image (0-bad index): contains left, top, right, bottom pixel coordinates
卑劣的意思3 dimensions 3D object dimensions: height, width, length (in meters) 3 location 3D object location x,y,z in camera coordinates (in meters) 1 rotation_y Rotation ry around Y-axis in camera coordinates [-pi..pi] 1 score Only for results: Float, indicating confidence in
detection, needed for p/r curves, higher is better.
附 转换代码：
import os
import numpy as np
import pandas as pd
import os.path as osp
def replace(file, old_content, new_content):
content =read_file(file)
content = place(old_content, new_content)
rewrite_file(file, content)
# 读⽂件内容
def read_file(file):
with open(file, encoding='UTF-8') as f:
read_all = f.read()
f.clo()
return read_all
# 写内容到⽂件
def rewrite_file(file, data):
with open(file,'w', encoding='UTF-8') as f:
f.write(data)
f.clo()
src_data='/media/ckq/data/kitti/MOT/images/train'
qs =[s for s in os.listdir(src_data)]
#print(qs)
for q in qs:
path=osp.join(src_data,q,'')
# q_gt_path = osp.join(src_data, q, '')
# print(q_gt_path)
# gt = np.loadtxt(q_gt_path, dtype=np.str, delimiter=',') # 加载成np格式
# print(str(gt))
replace(path,' ',',')
replace(path,'DontCare','10')幼儿园指南
replace(path,'Person','1')
replace(path,'Pedestrian','2')
replace(path,'Car','3')
replace(path,'Person_sitting','4')
replace(path,'Cyclist','5')
replace(path,'Van','6')
replace(path,'Truck','7')
replace(path,'Tram','8')
replace(path,'Misc','9')
2.之后开始给每个数据及的每张图⽚进⾏标签：⾸先说⼀下我的需求是标注Car： kitti数据集标签⽣成 gen_lables_kitti_car.py：
附代码（对照FairMOT 和`FairVehicle的⽣成标签代码写）：
import os.path as osp
import os
import shutil
import numpy as np
def mkdirs(d):
# if ists(d):
if not osp.isdir(d):
os.makedirs(d)
data_root ='/media/ckq/data/kitti/'
q_root = data_root +'MOT/images/train'
label_root = data_root +'MOT/labels_with_ids/train'
label_root = data_root +'MOT/labels_with_ids/train'
if not os.path.isdir(label_root):
mkdirs(label_root)
el: # 如果之前已经⽣成过：递归删除⽬录和⽂件，重新⽣成⽬录
<(label_root)
os.makedirs(label_root)
cls_map ={
'Person=1'
'Pedestrian=2'
'Car=3'摄相
'Person_sitting=4'
'Cyclist=5'
'Van=6'
'Truck=7'
'Tram=8'
'Misc=9'
'DontCare=10'
}
print("Dir %s made"% label_root)
#qs = [s for s in os.listdir(q_root)]
#qs=['0000']
qs=['0000','0001','0002','0003',
'0004','0005','0006','0007',
'0008','0009','0010','0011',
'0012','0014','0015','0018',
'0019','0020']
#打印序列
print(qs)
tid_curr =0
tid_last =-1
total_track_id_num =0
for q in qs: # 每段视频都对应⼀个gt.txt
print("Process %s, "% q, end='')
q_info_path = osp.join(q_root, q,'qinfo.ini') #提取每个数据的info信息/media/ckq/data/kitti/MOT/images/train #print(q_info_path)
with open(q_info_path) as q_info_h: # 读取*.ini ⽂件
q_info = q_ad()
q_width =int(q_info[q_info.find('imWidth=')+8:q_info.find('\nimHeight')]) # 视频的宽
q_height =int(q_info[q_info.find('imHeight=')+9:q_info.find('\nimExt')]) # 视频的⾼#print('q_width:',q_width)
#print('q_height:', q_height)
gt_txt = osp.join(q_root, q,'gt','gt.txt') # 读取GT⽂件
#print(gt_txt) #打印路径
#gt = np.loadtxt(gt_txt, dtype=np.str, delimiter=',') # 加载成np格式
gt = np.loadtxt(gt_txt, dtype=np.float64, delimiter=',') # 加载成np格式
# print(gt) #打印⽂本内容
# print('gt.T')
# print(gt.T) #也是打印⽂本内容
idx = np.lexsort(gt.T[:2,:]) # 优先按照track id排序(对视频帧进⾏排序,⽽后对轨迹ID进⾏排序)
# print(idx)
gt = gt[idx,:]
tr_ids =t(gt[:,1])
print("%d track ids in q %s"%(len(tr_ids), q))
total_track_id_num +=len(tr_ids) # track id统计数量如何正确计算？
q_label_root = osp.join(label_root, q,'img1')
mkdirs(q_label_root)
# 读取GT数据的每⼀⾏(⼀⾏即⼀条数据)
# for fid, tid, x, y, w, h, mark, cls, vis_ratio in gt:
for fid, tid, type, truncated, occluded, alpha, \
忆西湖
bbox_left, bbox_top, bbox_right ,bbox_bottom ,_,_,_,_,_,_,_ in gt: #height, width, length , location_x,location_y,location_z , rotation_y in gt: # frame_id, track_id, top, left, width, height, mark, class, visibility ratio
#if cls != 3: # 我们需要Car的标注数据
if type !=3: # 我们需要Car的标注数据
continue
网球比赛计分规则# if mark == 0: # mark为0时忽略(不在当前帧的考虑范围)
# continue
# if vis_ratio <= 0.2:
# continue
fid =int(fid)
tid =int(tid)
# 判断是否是同⼀个track，记录上⼀个track和当前track
if not tid == tid_last: # not的优先级⽐==⾼
tid_curr +=1
tid_last = tid
#由于kitti标签与训练标签参数有点不同需要⾃⼰计算 x y w h
不完美的小孩歌词
w=float(bbox_right-bbox_left)
h=float(bbox_bottom-bbox_top)
x=int(bbox_left+0.5)
y=int(bbox_top+0.5)
# bbox中⼼点坐标
x += w /2
y += h /2
# ⽹label中写⼊track id, bbox中⼼点坐标和宽⾼(归⼀化到0~1)
# 第⼀列的0是默认只对⼀种类别进⾏多⽬标检测跟踪(0是类别)
label_str ='0 {:d} {:.6f} {:.6f} {:.6f} {:.6f}\n'.format(
tid_curr,
x / q_width, # center_x
y / q_height, # center_y
w / q_width, # bbox_w
h / q_height) # bbox_h
# print(label_str.strip())
发财树养殖
label_f_path = osp.join(q_label_root,'{:06d}.txt'.format(fid))
with open(label_f_path,'a') as f: # 以追加的⽅式添加每⼀帧的label
f.write(label_str)
print("Total %d track ids in this datat"% total_track_id_num)
简单又好做的手工print('Done')
kitti数据集标签⽣成 gen_lables_kitti_car.py：(Car Van Truck)
import os.path as osp
import os
import shutil
import numpy as np
def mkdirs(d):
# if ists(d):
if not osp.isdir(d):
