Convolutional Tables Ensemble: classification in microseconds
This addresses the need for fast, accurate object recognition in real-time applications, though it is incremental as it builds on existing fast architectures with optimizations.
The paper tackled the problem of object category recognition under severe time constraints (1-1000 microseconds) by proposing Convolutional Tables Ensemble (CTE), achieving accuracy improvements of 24-45% over related methods of similar speed and showing CTE can outperform CNNs in certain speed ranges or provide 5-200X speedup with similar error rates.
We study classifiers operating under severe classification time constraints, corresponding to 1-1000 CPU microseconds, using Convolutional Tables Ensemble (CTE), an inherently fast architecture for object category recognition. The architecture is based on convolutionally-applied sparse feature extraction, using trees or ferns, and a linear voting layer. Several structure and optimization variants are considered, including novel decision functions, tree learning algorithm, and distillation from CNN to CTE architecture. Accuracy improvements of 24-45% over related art of similar speed are demonstrated on standard object recognition benchmarks. Using Pareto speed-accuracy curves, we show that CTE can provide better accuracy than Convolutional Neural Networks (CNN) for a certain range of classification time constraints, or alternatively provide similar error rates with 5-200X speedup.