Processed cuhk-sysu has 11.7k id's but original cuhk-sysu has 8.4k id's ?

Zhongdao / Towards-Realtime-MOT

Joint Detection and Embedding for fast multi-object tracking

MIT License

2.38k stars 539 forks source link

Processed cuhk-sysu has 11.7k id's but original cuhk-sysu has 8.4k id's ? #115

Open sachingoyal0104 opened 4 years ago

Zhongdao commented 4 years ago

This is because I have not re-ordered the ID indices. Some ID index may be skipped. This may affect the softmax classifier somehow but the impact is supposed to be small.

Kewenjing1020 commented 4 years ago

I downloaded the dataset and it contains 11934 persons in the annotation. Why's there a gap?

Kewenjing1020 commented 4 years ago

Hi, how do interpret the dataset? Both in 'Person.mat' and 'Images.mat' , each bbox has a 0/1 label, what does this mean?

HanGuangXin commented 2 years ago

@Zhongdao Hi, can you share the code for converting CUHKSYSU dataset format?