Just after CVPR 2015, I described the exponential growth that deep learning had in vision paper (see the original blog post) by analyzing the paper titles.
Since the CVF just released the papers here, I updated the script and run it.
You can also find all papers in BibTex format in this folder.
Here the result in the command line:
And the plot:
Surprisingly, the percentage of papers with some of the deep-learning keywords is the same for ICCV 2015 than for CVPR 2015. Does this mean that deep learning has plateaued in computer vision?
I feel that the general opinion will be that the field is still far from reaching a plateau. My interpretation for the titles not increasing is that deep learning is almost taken for granted in some sub-fields, so some authors do not feel the need to specify it in the title.