Project Performance

Overall our project performed well. We have successfully created a system that can use the Sawyer wrist camera to scan a drawing on a piece of paper from a variety of angles, skeletonize the drawing into line art, translate this line art to traceable contours, plan an optimized path for drawing these contours, and draw the image on another arbitrarily placed piece of paper located using AR tags and the Sawyer head camera. In comparing the hand drawing to the robot drawing we found that the robot could effectively recreate the source drawing on a skewed output paper. However, within this drawing there were some imperfections. Some details of the drawing were lost, and some artifacts in the lines appeared, in addition to relatively jittery lines. The robot is also relatively slow in completing the drawings, since we left the speed at a slow rate (so that in case things went bad, we would have more time to react and hit the emergency stop button).

We found that Robo Artist works very well for simple to moderately complex drawings. These were replicated very accurately and in a reasonable amount of time. For more complex images, like detailed line art of a face, issues arose. The biggest issue is the low source image capture quality from the wrist camera. We tried different parameters in the image processing, but small details lost due to low capture resolution cannot simply be regenerated. In our overview video, for example, you can notice there were some artifacts in the output image that can be attributed to graininess and imprecise focus in the source image. These are some things that could be improved in the future with more work but we were successfully able to accomplish our base goals of creating a system that can scan a drawing and replicate it on another piece of paper.