hustvl / YOLOP

You Only Look Once for Panopitic Driving Perception.(MIR2022)
MIT License
1.94k stars 413 forks source link

detection experiment for BDD100K(Detection of classification) #91

Open superXR opened 2 years ago

superXR commented 2 years ago

Hi, I'd like to ask you some questions about your paper. Is the detection experiment in this paper only for vehicles? Do you have 10 or 13 categories of object detection experiments for data set BDD100K? What are their recall, accuracy and mAP50 values?

happypostgraduate123 commented 2 years ago

我也想问这个问题,在运行代码后跑出来的结果仅仅只有car的检测结果,没有行人等其他类别的检测结果,我在代码中的train.py中的第252行看到model.nc=1,请问这里的nc是否是指代物体种类(num_class),但我在论文中未看到有说在object detection任务里只检测car这一类

happypostgraduate123 commented 2 years ago

github在nc那一行不能把我的话原原本本的展示,希望不会妨碍大家理解我原始的意思

我想问这个,在运行代码后跑出来的结果只有第255个问题的检测结果,没有行人等其他类别的检测结果,我在代码中看到了train.py中的model.nc= 1、请问这里的nc是否是指代未分类的一类(num_class),但我在论文中看到有说在object detection任务里只检测car

whiterAutumn commented 2 years ago

As far as I know, it consider the "car, truck, bus, train' four class as the unique detection category 'car".

superXR commented 2 years ago

As far as I know, it consider the "car, truck, bus, train' four class as the unique detection category 'car".

I agree with you.

thinkthinking commented 2 years ago

我也想问这个问题,在运行代码后跑出来的结果仅仅只有car的检测结果,没有行人等其他类别的检测结果,我在代码中的train.py中的第252行看到model.nc=1,请问这里的nc是否是指代物体种类(num_class),但我在论文中未看到有说在object detection任务里只检测car这一类

是的,nc是number of classes的缩写, 在commom.py 178行有注释。