search

dk-liang / FIDTM

[IEEE TMM] Focal Inverse Distance Transform Maps for Crowd Localization
MIT License
169 stars 41 forks source link

RDTM is similar to Inverse k-Nearest Neighbor Maps #2

Closed HartLen closed 3 years ago

HartLen commented 3 years ago

I think the proposed RDTM is almost the same with Inverse k-Nearest Neighbor Maps in [1], except the name and experiments for localization.

[1] Improving Dense Crowd Counting Convolutional Neural Networks using Inverse k-Nearest Neighbor Maps and Multiscale Upsampling

dk-liang commented 3 years ago

I think the proposed RDTM is almost the same with Inverse k-Nearest Neighbor Maps in [1], except the name and experiments for localization.

[1] Improving Dense Crowd Counting Convolutional Neural Networks using Inverse k-Nearest Neighbor Maps and Multiscale Upsampling

Thanks for your attention. We do not think the RDTM is the same as the IKNN map, there are two crucial reasons: 1) IKNN map is a KNN distance map instead of a distance transform map. In other words, k-Nearest Neighbor is not the same as Distance Transform. In particular, R-DT is a special form of IKNN (when the K is set as 1). 2) The VISAPP paper is only a counting task instead of a localization task, and the localization task is one of our main contributions. To the best of our knowledge, we are the first to leverage the local maxima of the R-DT map for localization-based crowd counting. We will add discussion in the next version.