SHHA train
yanfangfangfang opened this issue · comments
result in the paper of shanghaitech partA is 64.8/107.5 can you give me the code of this part or some setting of your experiment .i have run the code of C-3 resnet50 but i got partA 77.3/124.5 with batchsize=6
crop image size of batch min size
I cannot reproduce the result for SHTA either. When I try to unsupervised case, I can only get MAE of ~230 if I train the model only with the filtered synthetic data (~2000 images). I have a big performance gap compared to the reported result (MAE: 160). Do you use different learning parameters?
with the batch size=2, deal with different size in two ways,1,crop to minsize,2 expand to maxsize but the result was not change much .the result in the paper of [Learning from Synthetic Data for Crowd Counting in the Wild(cvpr2019)]SHTA:64.8/107.5
.i reproduce this part also with pre-train modle on GCC,but got worse result.can you release your code about this part