mzolfaghari / ECO-efficient-video-understanding

Code and models of paper " ECO: Efficient Convolutional Network for Online Video Understanding", ECCV 2018
MIT License
437 stars 96 forks source link

setting the number of segments #47

Open wasahaiah opened 5 years ago

wasahaiah commented 5 years ago

The presented method setting the number of segments seems not right.

layer { name: "reshape_fc_st2" type: "Reshape" bottom: "global_pool2D" top: "reshape_fc_st2" reshape_param { shape { dim: [-1, 1, 16, 1024] } } }

layer { name: "segment_consensus_st2" type: "Pooling" bottom: "reshape_fc_st2" top: "pool_fusion_st2" pooling_param { pool: AVE kernel_h: 16 kernel_w: 1 } }

it looks like that the parameters in the above two layers need to be changed accordingly.