mots_tools

mots_tools copied to clipboard

I have an issue while converting MOTS(Multiple object tracking segmentation) ground truth values to standard MOT(Multiple object tracking) formats. **MOTS Format** For example:1 2002 2 1080 1920 UkU\1`0RQ1>PoN\OVP1X1F=I3oSOTNlg0U2lWOVNng0m1nWOWNlg0n1PXOWNlg0l1SXOUNjg0P2RXO.... Frame: 1,...

ashwath1295

"To check that no frame contains overlapping masks" : Can we use mots_tools for overlapping masks?

2

Hi, following your code I found that 1. In mots_common.io => load_txt(path) You have to check that no frame contains overlapping masks... if frame not in combined_mask_per_frame: combined_mask_per_frame[frame] = mask...

PanaddaK

mots_common/io.py does not support folder structure for MOTS dataset?

3

If one downloads the MOTS dataset from https://motchallenge.net/data/MOTS/, you get a folder structure for the training set, where ground truth files are all called `gt.txt` and placed in different folders...

ahrnbom

Incorrect seqmap for the actual train set in MOTS?

1

On https://motchallenge.net/data/MOTS/ the MOTS training set consists of the four videos MOTS20-02, MOTS20-05, MOTS20-09 and MOTS20-11. The seqmap corresponding to this one is called `val_MOTSchallenge.seqmap`. Shouldn't this be the one...

ahrnbom

I tryed to use _mask.py_ to decode annotation but failed,and at the same time, I also use the `load_txt()`in _mots_tools_，which failed again.Can you give me some advice?

AzulYang