search

aleksispi / drl-rpn-tf

Official Tensorflow implementation of drl-RPN: Deep Reinforcement Learning of Region Proposal Networks (CVPR 2018 paper)
MIT License
77 stars 21 forks source link

not an issue - question on line detection research #1

Closed johndpope closed 6 years ago

johndpope commented 6 years ago

are you aware of any work to do simple line detection using ML? the hough transform is terribly expensive and i'd like to fine a cheap way to do in real time for mobile. https://en.wikipedia.org/wiki/Hough_transform This is usually done with canny line detection. but I have a suspicious losing the colors is throwing away potentially useful information.

There's is Parallel Coordinate lines — Line detection http://medusa.fit.vutbr.cz/public/data/papers/2011-CVPR-Dubska-PClines.pdf

https://github.com/syagev/deep-ellipse/issues/1

if the nets could be trained to find the lines / then this could be put into a pipeline of further computations to find perspectives / walls etc.

aleksispi commented 6 years ago

Hi! Sorry, but I don't have any such experience. There are fast detectors such as yolo v3 you could check out.