wonder Expert Licensed User Longtime User Jun 10, 2019 #1 Hi! Having seen that there are a few OpenCV / EdgeDetection libraries around, I would like to ask their creators/users if it's possible to accomplish the task of auto detecting comic book / manga panels. Note 1: It should also work for grayscale and line-art (black and white) input sources. Note 2: There are non-standard panel shapes like triangles, trapezoids and circles. Note 3: If someone proposes a custom solution, I will gladly pay for it (as much as I can afford). Page.jpg (input): Output:
Hi! Having seen that there are a few OpenCV / EdgeDetection libraries around, I would like to ask their creators/users if it's possible to accomplish the task of auto detecting comic book / manga panels. Note 1: It should also work for grayscale and line-art (black and white) input sources. Note 2: There are non-standard panel shapes like triangles, trapezoids and circles. Note 3: If someone proposes a custom solution, I will gladly pay for it (as much as I can afford). Page.jpg (input): Output:
JordiCP Expert Licensed User Longtime User Jun 12, 2019 #2 Just seen it (not very active here lately). Not sure if still interested, but shouldn't be too difficult with OpenCV if the background behind the panels is a solid color Upvote 0
Just seen it (not very active here lately). Not sure if still interested, but shouldn't be too difficult with OpenCV if the background behind the panels is a solid color
wonder Expert Licensed User Longtime User Jun 13, 2019 #3 Yes I am! Indeed, the background is usually either solid black or white. There are some odd shapes every now and then though, also "panel-in-panel", which might be trickier handle. I was wondering if OpenCV by itself would be enough or if perhaps, this is more of a job for a Neural Network... I've never used the OpenCV lib before, by the way. How's the learning curve? Upvote 0
Yes I am! Indeed, the background is usually either solid black or white. There are some odd shapes every now and then though, also "panel-in-panel", which might be trickier handle. I was wondering if OpenCV by itself would be enough or if perhaps, this is more of a job for a Neural Network... I've never used the OpenCV lib before, by the way. How's the learning curve?
