Given the following 3 research papers, the authors have shown different heatmap graphical representations for features of the trained CNN models:
On the performance of Convnet feature for place recognition: link
NetVLAD: CNN architecture for weakly supervised place recognition: link
- Deep Learning Features at Scale for Visual Place Recognition: link
Does anyone know the easiest heatmap implementation in Python given deploy.protxt
and model_wights_bias.caffemodel
files?
PS: I am aware of the following answers and packages: answer1, package1 but they do not provide these solutions shown in figures above!
Thanks,