I am asking in vision category because it’s object recognition I’m after.
I am using a character based CNN for categorizing data strings. Inside those strings are sub-segments that I want to categorize, or recognize. I see it like the 2d object recognition in RCNN or SDD. They put bounding boxes around multiple objects in an image. That’s what I want to do: underline or tag substrings and identify their categories.
I also think my problem is analogous to 1d data like ECG data with the pqrst parts of the heart beat. Many of the ECG algorithms talk about identifying the unusual heartbeat but I haven’t found examples of putting boxes around the sub segment parts of the heartbeat (pqrst parts).
Anyone have any tips or direction to suggest on how I can change my conv1d character based CNN into a conv1d R-CNN or SDD.