Skip to content

Commit 1d2e3b4

Browse files
committed
📝 Update DATASET.md.
1 parent df1eb5d commit 1d2e3b4

File tree

1 file changed

+12
-12
lines changed

1 file changed

+12
-12
lines changed

docs/DATASET.md

Lines changed: 12 additions & 12 deletions
Original file line numberDiff line numberDiff line change
@@ -4,15 +4,15 @@
44

55
- [DanceTrack](https://github.com/DanceTrack/DanceTrack)
66
- [SportsMOT](https://github.com/MCG-NJU/SportsMOT)
7-
- [MOT17](https://motchallenge.net/data/MOT17/)
7+
- [BFT](https://george-zhuang.github.io/nettrack/)
88
- [CrowdHuman](https://www.crowdhuman.org/)
99

1010
## Generate GT files
1111

12-
For the MOT17 and CrowdHuman datasets, you’ll need to use the provided script to convert their ground truth files to the format we require:
12+
For the BFT and CrowdHuman datasets, you’ll need to use the provided script to convert their ground truth files to the format we require:
1313

14-
- For MOT17: [gen_mot17_gts.py](../data/gen_mot17_gts.py)
15-
- For CrowdHuman: [gen_crowdhuman_gts.py](../data/gen_crowdhuman_gts.py)
14+
- For BFT: [gen_bft_gts.py](../tools/gen_bft_gts.py)
15+
- For CrowdHuman: [gen_crowdhuman_gts.py](../tools/gen_crowdhuman_gts.py)
1616

1717
:pushpin: You need to modify the paths in the script according to your requirements.
1818

@@ -34,12 +34,14 @@ For the MOT17 and CrowdHuman datasets, you’ll need to use the provided script
3434
│ ├── train_seqmap.txt
3535
│ ├── val_seqmap.txt
3636
│ └── test_seqmap.txt
37-
├── MOT17/
38-
│ ├── images/
39-
│ │ ├── train/ # unzip from MOT17, you need to delete some files, see the below the instructions
40-
│ │ └── test/ # unzip from MOT17
41-
│ └── gts/
42-
│ └── train/ # generate by ./data/gen_mot17_gts.py
37+
├── BFT/
38+
│ ├── train/
39+
│ ├── val/
40+
│ ├── test/
41+
│ ├── annotations_mot/ # used for generate gts for BFT
42+
│ ├── train_seqmap.txt
43+
│ ├── val_seqmap.txt
44+
│ └── test_seqmap.txt
4345
└── CrowdHuman/
4446
├── images/
4547
│ ├── train/ # unzip from CrowdHuman
@@ -49,8 +51,6 @@ For the MOT17 and CrowdHuman datasets, you’ll need to use the provided script
4951
└── val/ # generate by ./data/gen_crowdhuman_gts.py
5052
```
5153

52-
:warning: Since each video sequence in MOT17 is stored three times, for each training video sequence, you should delete the other two sequences to achieve deduplication. For instance, in my experiment, I only retained the ‘xxx-DPM’ sequences.”
53-
5454
## Q & A
5555

5656
- Q: Lack the `val_seqmap.txt` file of SportsMOT? </br>

0 commit comments

Comments
 (0)