import configparser
import csv
import glob
import os
# from pathlib import Path
import fiftyone as fo
[docs]def read_detections(sample, fname):
"""
:param sample: fiftyone.Sample
:param fname: frame-by-frame annotations
TVD format
::
[Frame_Index, Object_ID, Top_left_x, Top_left_y, Width, Height, Confidence, 3D_x, 3D_y] ?
or is one of them a class label..? what label set?
note that these are abs coordinates
Example:
::
1,1,193,686,125,331,1,1,1
2,1,193,686,124,330,1,1,1
3,1,194,686,124,330,1,1,1
4,1,197,684,116,339,1,1,1
5,1,194,684,121,330,1,1,1
6,1,199,685,113,335,1,1,1
...
543,1,645,855,47,125,1,1,1
544,1,646,860,48,118,1,1,1
1,3,746,894,1098,106,0,9,1
2,3,746,894,1098,106,0,9,1
...
i.e. note that frame indexes can start again from 1
"""
with open(fname, "r") as csvfile: # per frame
dets_per_frame = {}
for line in csv.reader(csvfile, delimiter=","): # per detection
(
Frame_Index,
Object_ID,
Top_left_x,
Top_left_y,
Width,
Height,
Confidence,
_,
_,
) = line
Frame_Index = int(Frame_Index)
object_id = int(Object_ID)
x0 = float(Top_left_x) / sample.metadata.frame_width
y0 = float(Top_left_y) / sample.metadata.frame_height
w = float(Width) / sample.metadata.frame_width
h = float(Height) / sample.metadata.frame_height
confidence = float(Confidence)
# bbox center coords:
# x0=x0-w/2
# y0=y0-h/2
# label=classmap[n_class]
bbox = [x0, y0, w, h]
if Frame_Index not in dets_per_frame:
dets_per_frame[Frame_Index] = []
dets_per_frame[Frame_Index].append(
fo.Detection(
# label=label, # no label? # for MOT we can always substitute the label with the id..!
label="#" + str(object_id),
# index=object_id,
confidence=confidence,
bounding_box=bbox,
)
)
for Frame_Index, dets in dets_per_frame.items():
detections = fo.Detections(detections=dets)
f = fo.Frame(detections=detections)
sample.frames.add_frame(frame=f, frame_number=Frame_Index)
[docs]def register(dirname, name="tvd-object-tracking-v1"):
"""Register tencent video dataset (TVD), object tracking subset.
The directory structure for this looks like:
::
dirname/
|
├── TVD-01
│ ├── gt
│ │ └── gt.txt
│ ├── img1
│ └── seqinfo.ini
├── TVD-01.mp4
├── TVD-02
│ ├── gt
│ │ └── gt.txt
│ ├── img1
│ └── seqinfo.ini
├── TVD-02.mp4
├── TVD-03
│ ├── gt
│ │ └── gt.txt
│ ├── img1
│ └── seqinfo.ini
└── TVD-03.mp4
"""
dirs = os.path.join(dirname, os.path.join("TVD-*", "")) # in order to get dirs only
print("searching for", dirs)
dirlist = glob.glob(dirs)
if len(dirlist) < 1:
print("no directories found, will exit. Check your path")
return -1
# print(dirlist)
if name in fo.list_datasets():
print("Dataset", name, "exists. Will remove it first")
fo.delete_dataset(name)
dataset = fo.Dataset(name)
print("Dataset", name, "created")
for dir_ in dirlist:
# /path/to/TVD-01/
print("\nIn directory", dir_)
tag = dir_.split(os.path.sep)[-2]
print(tag) # TVD-01
gt_file = os.path.join(dir_, "gt", "gt.txt")
print(gt_file)
assert os.path.isfile(gt_file), "can't find " + gt_file
video_file = os.path.join(dirname, tag + ".mp4")
print(video_file)
assert os.path.isfile(video_file), "can't find " + video_file
ini_file = os.path.join(dirname, tag, "seqinfo.ini")
print(ini_file)
assert os.path.isfile(ini_file), "can't find " + ini_file
config = configparser.ConfigParser()
config.read(ini_file)
seq = config["Sequence"]
# frameRate = int(seq["frameRate"])
imWidth = int(seq["imWidth"])
imHeight = int(seq["imHeight"])
metadata = fo.VideoMetadata(frame_width=imWidth, frame_height=imHeight)
video_sample = fo.Sample(
media_type="video", filepath=video_file, metadata=metadata, name_tag=tag
)
print("importing detections for", video_file, "from", gt_file)
read_detections(video_sample, gt_file)
# video_sample now has the detections
dataset.add_sample(video_sample)
dataset.persistent = True
print("\nDONE!")
