Description: Thermal images are captured during the fabrication of Aerospace composite structures. These images can provide useful insight to the quality of the part during fabrication. Detection of flaws during the automatic fiber placement can reduce cost by preventing repair or rebuild. We are developing an algorithm for automatically detecting common flaws and anomalies to reduce the time needed to inspect the plies during fabrication. This requires a great deal of labeled data for the training of a CNN.
This task was to label data frames for machine learning.