Consideration of "Local Importance-based Pooling" (LIP) as an attention method

[sebgao]

Hi Menghao,

Thanks for your great and helpful work. I regard our proposed LIP [1] also as a local attention method designed for the spatial downsampling procedure, whose weights are produced by a lightweight sub-network. Though being designed for spatial downsizing, it still falls in the regime of attention approaches. Could you please consider LIP also as an attention method especially for the pooling procedure?

By the way, the work TAM [2] done by our group is accepted by ICCV 2021. Please consider updating the venue of this paper.

[1] LIP: Local Importance-Based Pooling. ICCV 2019.
[2] TAM: Temporal Adaptive Module for Video Recognition. ICCV 2021.

[MenghaoGuo]

Hi,

Thanks for your great work and kind suggestion for perfecting this paper. I will read the mentioned papers and consider your suggestions.

Best,
Menghao