Two-month plan
Eight weeks, with a verifiable result at the end of each week.
| Week | What I'll learn | Verifiable result |
|---|---|---|
| Week 1 | Laser triangulation principles, the LJ-X8000 manual, a Python environment and repository | A one-page principles note and commands that rebuild the environment |
| Week 2 | Connect to the controller, read and save single and continuous profiles | 100 consecutive profiles of one object, with a profile plot and noise statistics |
| Week 3 | Three-axis controller communication, single-axis motion, homing and limits | A script that moves to a target position and reports it, with safety limits set |
| Week 4 | Constant-speed scanning, aligning profiles with positions (encoder trigger or timestamps) | A scan of a standard part producing the first height map |
| Week 5 | Invalid points and outliers, plane-fit leveling | Before-and-after leveling plots and the measured error on a known step height |
| Week 6 | First defect detection: baseline comparison, thresholds, morphology, connected components | Defects marked on a deliberately flawed sample, with a report |
| Week 7 | Repeated scans, evaluation metrics (detection rate, false alarms), fixed configs and seeds | The same command reproduces results, plus an evaluation table |
| Week 8 | Integrate into a command-line flow, prepare thesis figures and a README | One command from scan to defect report, with a demo record |
Each week also gets a little time for the investing and education side projects, to practice data handling and agent design. If a week falls behind, I protect the verifiable result first and adjust later weeks.