A look at 2000s speech synthesis: how unit selection and HMM-based voice models worked, why they sounded different, and what listening tests revealed.
How Conferences Shaped AI Research in the 2000s
A look at how 2000s AI conferences shaped benchmarks, methods and publication practices across speech, vision and machine learning.
Object Tracking in the 2000s: Motion, Occlusion, and Identity
How 2000s visual trackers predicted motion, handled occlusion, matched detections, and measured identity errors across video frames.
What Small Speech Corpora Could—and Couldn’t—Show in the 2000s
How 2000s speech researchers built and tested recognizers for under-resourced languages, from acoustic transfer and lexicons to careful evaluation.
Machine Learning in 2000s Recognition Research: Features, Models, and Evaluation
How 2000s researchers combined designed features and learned models in speech, vision, and document analysis—and tested whether their results generalized.
How AI Conferences Shaped Research in the 2000s
How 2000s AI conferences helped researchers compare results, share methods across fields, and expose the limits of benchmarks and published papers.
Beyond OCR: How 2000s Systems Learned to Read Page Structure
How mid-2000s document analysis moved beyond OCR to handle page layout, reading order, tables, forms and verifiable extraction.
Reading Mid-2000s Stereo Vision Research
How mid-2000s stereo papers balanced matching costs, smoothness, and occlusions—and what Middlebury scores could and could not show.
Biometrics in the 2000s: Beyond the Match Score
How capture quality, templates, benchmarks, and large-scale deployment changed biometric research in the 2000s.
Medical Imaging AI in the 2000s: What CAD Could—and Couldn’t—Do
A look at 2000s medical imaging AI, from mammography CAD and lung nodule detection to scan measurements, clinical testing, and false alarms.