search

xiaohaoChen / rrc_detection

Accurate Single Stage Detector Using Recurrent Rolling Convolution
Other
361 stars 136 forks source link

how about pedestrian parameters for training? #35

Open ChenLi2013 opened 7 years ago

ChenLi2013 commented 7 years ago

could you show your train parameters for pedestrian,cyclist, i want to use rrc to get results showed by your paper ,could you give me some suggestions?

jaminChen999 commented 7 years ago

hi，@ChenLi2013，i can't get the results of pedestrian,cyclist showed in the paper by only kitti's datasets,And you ?

ChenLi2013 commented 7 years ago

@jaminChen999 ,i also can't get the results from kitti's datasets for testing pedestrian and cyclist .

XiongweiWu commented 6 years ago

@ChenLi2013 @jaminChen999 I am highly confused with the results reported in paper. For example, whether they have applied ensemble models in pedestrian or cyclist detection? and how they get the results on Car hard case, also by ensemble or not? since the given model performs much worse than expectation(~80% mAP vs ~86% paper said).