xhp-hust-2018-2011 / SS-DCNet

From Open Set to Closed Set: Supervised Spatial Divide-and-Conquer for Object Counting
78 stars 19 forks source link

Results better than reported in the paper? #12

Open svena33 opened 4 years ago

svena33 commented 4 years ago

In the paper (and in the first table in the Readme.md) you report: SS-DCNet (cls) | 56.1 (SHA) | 6.6 (SHB)

When running the evaluation code on SHA and SHB I find the following MAE results: SS-DCNet | 55.571 (SHA) | 6.645 (SHB)

These are the same results as you report a bit later in the Readme.md.

My question is hence the following: Is the model provided here more performant on the Shanghai dataset part A than the reported SotA model in your paper?