ICA SUCCEEDS
- ICA achieves an error rate of 34.59% using
RGB + Long Lines + Entropy as features that is comparable to the
average error performance using intelligent guessing which is only
33.3%.
- The error rate is well below the error performance that was
achieved using the other classification algorithms discussed above.
- No confusion between the LSBE and HSBE class. This was a
predominant shortcoming of the other classification algorithms
discussed above.
- Shown below is the confusion matrix that was obtained as a
result of testing on set 01 using rgb, long lines and entropy as
feature vectors. The error rate for this experiment is 34.59%.