Shane Wighton had to make a few adjustments to Jigzilla to tackle an incredibly tough 4,000 piece all-white puzzle and after seeing it in-action, you’ll agree that the upgrades were sufficient. The most notable of the upgrades was completely reprogramming its AI algorithm, which was necessary since there are no images on the pieces and could take a human years to finish.
The training process involved training the robot using photographs of every puzzle piece using a telecentric lens, which took 8-hours. Next, the shape of each piece had to be extracted from the photos before the preprocessing could begin. Preprocessing was another nightmare that took another few hours to complete on a Puget Systems computer. Lastly, software optimizations made Wighton realize that the original pieces all had rough edges, resulting him having to make his own puzzle from scratch.
- Official product by Beverly
- Makes a great gift!
- Cute and Collectible
