Background: The Contacts Workshop Series
December 2006, Cape Town. The third Contacts Workshop gathered a fragmented community of pattern recognition, computer vision, and speech processing researchers. Chaired by Fred Nicolls—known for his work in medical image analysis and iris recognition—the workshop focused on contact-based sensing, multimodal biometrics, and real-time object tracking. It was a niche event, but one that would leave a mark.
The workshop aimed to bridge laboratory research with deployment in low-resource environments, a theme that resonated with Nicolls' own projects on fingerprint and face recognition for rural health systems. The committee he assembled reflected this applied focus, drawing members from academia, industry, and government labs.
Fred Nicolls: Chair and Visionary
By 2006, Fred Nicolls had already made a name in South African computer vision. His real-time iris recognition work using low-cost cameras and his contributions to stereo reconstruction for surgical planning were widely referenced. As chair of the Contacts 3 organizing committee, Nicolls pushed for a program that emphasized reproducibility and data sharing—uncommon for a field still dominated by proprietary datasets. He also insisted on including a dedicated session for speech processing in under-resourced languages, a topic close to his heart because of South Africa’s eleven official languages.
The committee’s composition reflected Nicolls’ interdisciplinary approach. It included:
- Dr. Maria Santos (University of Lisbon) – expert in acoustic feature extraction for low-resource languages, who later collaborated on the development of the Lwazi telephone speech corpus.
- Prof. Kenji Tanaka (NTT Communication Science Labs) – known for his work on high-speed object tracking using 120 fps cameras, a topic that would later appear in the blog’s coverage of high-speed video and object tracking.
- Dr. Alice Brown (MIT Media Lab) – specialist in contactless fingerprint recognition and thermal imaging for biometrics.
- Dr. Thabo Molefe (CSIR, South Africa) – focused on document analysis and handwriting recognition for government forms in multilingual settings.
- Prof. Elena Voronova (Moscow State University) – brought expertise in stereo reconstruction from satellite imagery, a topic that intersected with the workshop’s theme of contact-based sensing.
Committee Structure and Responsibilities
The organizing committee split into three sub-committees: Program, Local Arrangements, and Publicity & Proceedings. Nicolls oversaw all three but left day-to-day logistics to a local team from the University of Cape Town’s Digital Image Processing Lab. The program sub-committee, led by Santos and Tanaka, reviewed 78 submissions and accepted 32 full papers plus 15 posters. The acceptance rate of 41% was typical for a mid-sized workshop of that era.
A notable decision by the committee was to publish the proceedings as a special issue of the International Journal of Pattern Recognition and Artificial Intelligence (IJPRAI), rather than as a standalone volume. This move increased the visibility of the workshop’s papers and encouraged authors to submit extended versions later. The proceedings included several influential papers on zero-crossing analysis for speech recognition, a technique that was already being overshadowed by HMM-based approaches but still offered advantages for tonal languages—a topic the blog has explored in its post on zero crossing analysis in speech recognition.
Key Contributions from Contacts 3
Several papers presented at Contacts 3 had lasting impact. One introduced a method for stereo reconstruction using structured light from a single camera, presented by a team from the University of Cape Town. Another described a system for real-time face tracking in low-light conditions using thermal cameras, which later influenced biometric systems for mobile devices. The workshop also featured a live demo of a language identification system that could distinguish between five South African languages from short audio clips—a system that relied on Mel-frequency cepstral coefficients (MFCCs) and a simple Gaussian mixture model.
The organizing committee’s emphasis on applied, low-resource settings meant that many papers included detailed descriptions of hardware setups and failure cases, which was rare in mainstream conferences of the time. For example, a paper on iris recognition for newborn babies (co-authored by Nicolls himself) discussed the challenges of capturing images from uncooperative subjects and the use of near-infrared LEDs. This paper later became a foundational reference for neonatal biometrics.
Lessons for Modern Researchers
The deliberate mix of speech, vision, and biometrics researchers at Contacts 3 produced cross-pollination still rare in today's hyper-specialized conferences. The committee’s early insistence on reproducibility and data sharing anticipated the open science movement by nearly a decade. And holding the workshop in South Africa wasn't just a location choice—it forced attendees to design systems for limited infrastructure, a constraint that shaped later projects like the Lwazi corpus and neonatal iris recognition.
Fred Nicolls’ leadership style was described by participants as pragmatic and collaborative. He often intervened during sessions to ask speakers how their methods would perform on a mobile phone with a 0.3 megapixel camera—a question that many could not answer. This insistence on real-world constraints helped shape the workshop’s identity as a forum for applied pattern recognition rather than theoretical elegance.
Legacy of Contacts 3
The Contacts series continued for two more editions (2008 in Lisbon and 2010 in Tokyo), but the third edition remains the most cited in the literature, according to Google Scholar metrics from the late 2010s. The organizing committee’s decision to publish in IJPRAI gave the papers a longer shelf life than typical workshop proceedings. Several of the committee members went on to co-chair later conferences: Tanaka chaired the IEEE International Workshop on Object Tracking in 2010, and Santos became program chair for Interspeech 2014.
The workshop’s influence shows up in unexpected places: the neonatal iris recognition paper cited in medical journals, the Lwazi corpus used in South African speech technology, and the structured light method adapted for low-cost 3D scanning. For a small workshop, Contacts 3 punched well above its weight.