04/28/2006 Project launching meeting (Dr. Tasi, Dr. Wang, Pilho
– namly All members): Discuss the topic of the project (MUTCD) for
Pilho’s summer RA.
05/04/2006 Regular meeting (all): The objective of this meeting
is to discuss with the preliminary thought that Pilho have on
developing the NN and Dr. Wang will provide Pilho with the input
ROIs.Continue reviews on existing approaches. Manually select tens
of warning sign candidates from MUTCD databases. Find proper image
abstractions or features for neural network training.
05/12/2006 Regular meeting (all): We watched the traffic warning
class signs together and discussed the next approaches to be
composed in a number of steps. Dr. Wang suggested to make a system
be modularized so that a neural network specifically for the
recognition part (expected outputs namly W-1, W-2, etc.) to be
independable in the view of implementation to priori systems like
feature extractors including correlation computing modules or OCR.
In summary, a neural network recognition module should use the
features extracted by heterogeneous computing modules that extract
ROI and associated features from a raw image in a unified way to
classify the input into one of predefined categories.
05/17/2006 Regular meeting (all): Extensive discussions were
done on the proper features to extract from the traffic sign raw
images. In addition, the proper neural architecture in which number
of dimensions is also talked about. In conclusion, we set the next
intermediate goal as testing the performance of line receptors as
image abstraction features and plus finding a proper ANN engine for
05/25/2006 Regular meeting (all): Discussion on the on-going
project that measures the location of buoys on the Savannah port.
ANN engine tests were reported and FANN is selected as our
preliminary ANN engine. We will keep extending ANN training
experiment for the real traffic sign cases and future enhancing its