results of other datasets

RQuispeC / pytorch-ACSCP

Unofficial implementation of "Crowd Counting via Adversarial Cross-Scale Consistency Pursuit" with pytorch - CVPR 2018

MIT License

18 stars 4 forks source link

results of other datasets #7

Closed EillotY closed 4 years ago

EillotY commented 4 years ago

use the original code in shanghai dataset part A has MAE:92 which is far from the author releases result.I dont know which part cause problem .I set --people-thr 20 and also use the data_increating.So i want to know have you get a closer result of MAE:75 of shanghai_partA. Thank you very much!

RQuispeC commented 4 years ago

Hi @EillotY I didn't test this code for shanghai-tech, however I think that using --people-thr 20 for that dataset may create too few samples for training as shanghai-tech's images have less people amount that ucf-cc-50, for shanghai-tech (both part A and B) I would suggest to use --people-thr 0. What's data_increating? didn't get that part Hope you can replicate results for shanghai-tech

EillotY commented 4 years ago

ok ,i will try it.thanks a lot.