In 2017 and 2018 I worked with an brilliant signal processing engineer to explore the option of light diffusion (surface), computer vision (edges, corners, centering), and a neural network built around the concept for grading sports cards.
It is very possible; however, was too time and resource consuming for me to bring the concept to reality while I worked on IIB and other endeavors. I still have the prototype including the server to run it sitting in a closet. The sample size of data needed to curate real and proper grades was immense and the system is only as good as the engineer and the grader educating the engineer to work on it. Needless to say, a proper education process would take one person years or a team of industry experts many months (and a far more complex system to mesh their grading opinions into the database). Even then, if something were done so well a human couldn't detect it, how could a human built system detect it? There are impossible to identify restoration jobs.
The barrier of entry into the market is the denial of the collectors holding billions of dollars in inventory that do not want to hear that what they have has been altered, even if that is the reality. It is, without a doubt, immense.
That being said... there are restoration jobs so perfect that they would fool any professional including a well trained system. If fragments of the Dead Sea Scrolls sold for $500,000,000 can be faked so well that they fool experts who spent months inspecting them for authenticity, are displayed in a museum exhibit, and was only found out after more than a dozen industry scholars inspected them and committed to, out of their own pocket, expensive tests to prove their opinion-- what makes you think some nefarious individual(s) with a mid to high level set of skills can't alter, or even create from scratch, a sports card? Don't fool yourself.
Last edited by Sean1125; 08-29-2019 at 12:33 PM.
|
08-29-2019, 12:32 PM
Similar Threads
Threaded Mode