Research Article
Multiple Feature Fusion Based on Co-Training Approach and Time Regularization for Place Classification in Wearable Video
Algorithm 1
The CO-DAS and CO-TA-DAS algorithms.
INPUT: | Training set ; | Testing set ; | OUTPUT: | —class estimations for the testing set ; | , —trained classifiers; | PROCEDURE: | Compute visual features for every image in the dataset | Initialize and | Initialize and | Create two work sets and | Repeat until the sets and are empty (CO) | (a) Train classifiers , using the sets , respectively; | (b) Classify the patterns in the sets and using the classifiers and respectively; | (i) Compute scores and confidences on the set | (ii) Compute scores and confidences on the set | (c) Add the top confidence estimations , | (i) | (ii) | (d) Remove the top confidence patterns from the working sets | (i) | (ii) | (e) Go to step . | Optionally: perform Temporal Accumulation (TA) according to (15) | Perform classifier output fusion (DAS) | (a) Compute fused scores ; | (b) Output class estimations from the fused scores |
|