About Event

Welcome to the 19th Biennial meeting of The International Association for Craniofacial Identification - Race and Face: bias in forensic and archaeological investigation. Register for this symposium now and follow this conference on social media using #RaceAndFaceIACI


IACI is an international group of professionals holding common interest in craniofacial identification, broadly defined as any undertaking to use skulls and/or faces for human identification. Find out more about IACI here This conference is hosted by Face Lab at Liverpool John Moores University.

If you would like to present a short paper or poster at this symposium, please submit an abstract with your name and affiliation to facelab@ljmu.ac.uk by 1 July 2021. Please forward this to any interested parties.



Welcome and Introduction: Welcome and introduction

Prof. Caroline Wilkinson
Director of the Face Lab and the Director of the School of Art & Design at Liverpool John Moores University.

Race Trouble and DNA-phenotyping Forensic genetic technologies have constituted a major change in criminal investigation. These technologies have been rightly acclaimed and viewed as constituting a golden standard in the realm of forensic research, and in identification. Yet, while forensic DNA has been heralded as the ultimate identifier of the individual suspect, a truth machine (Lynch et al 2008), it has also been haunted by the specter of race. This is even more so in the practice of DNA phenotyping and associated technologies. A practice of giving a face to an unknown person based on DNA. I will argue that this aim of individualizing, the giving of a face, goes hand in hand with the doing of race. This practice prompts a number of questions: is it possible to move beyond race, and if so how? If not, how to stay with the trouble of race, the potentiality of race and racialization, and how to use such technologies in a way that does not lead to racism and the criminalization of large groups of people?

Prof. Amade M’charek
Affiliation Professor of Anthropology of Science at the Department of Anthropology, University of Amsterdam

Does presence of a face covering affect a person’s ability to construct an identifiable facial composite? – It is known that concealing parts of a face can decrease recognition (e.g., Min, Hadid, & Dugelay, 2011), while removing a concealment can improve it (e.g., Dagnes, Vezzetti, Marcolin, & Tornincasa, 2018). The current research investigated the effectiveness of a facial composite constructed of a target seen with or without a face covering. Thirty-two participants were shown a face of unfamiliar footballer with or without presence of a face mask. After nominally 24 hours, each person constructed a single composite of the face using the EvoFIT composite system. Two additional sets of stimuli were then prepared, with a mask was added to composite faces constructed without one, and vice versa. A different set of participants recruited on the basis of being familiar with international-level football attempted to name one of these four sets of composites. Faces constructed without a mask were correctly named significantly less than those constructed with a mask (M = 30.0 vs. 36.7%), although the effect was only medium sized and so composites had good naming in spite of the presence of the face covering. Adding a mask to a composite that had been constructed without one, however, led to substantially worse naming (M = 18.3%), indicating the importance of context for face construction as well as how addition of a face mask to a finished composite can greatly interfere with recognition.

Tom Barnes
University of Central Lancashire
Prof. Charlie Frowd
University of Central Lancashire

Five minutes break


Super-recognisers and the cross-race effect? A large body of research evidence has found that in general people are better at recognising and simultaneously matching the faces of people of their own ethnicity, than those from other ethnicities. Greater contact and interest in individuals from other ethnicities may reduce the impact of this effect. Super-recognisers possess exceptional face recognition and matching ability and had achieved scores expected of approximately the top 2% of the population on two face recognition tests. In the first study, White super-recognisers (n = 35) significantly outperformed White controls (n = 360) on the Glasgow (White) Face Matching Test, the Models (White) Face Matching Test, and an Egyptian Face Matching Test, albeit individual analyses demonstrated that not all super-recognisers produced superior scores. In the second study, White Super-Recognisers (n = 45) outperformed White controls (n = 538) at correctly recognising that they had or had not seen a Black or White face before. Evidence was also found that increased contact with Black people, together with higher White face matching scores, reduced the impact of the Cross-Ethnicity Effect. This research supports the use of super-recognisers in identification-critical job roles regardless of likely target ethnicity.

Prof. Josh Davis
Face and Voice Recognition Lab, University of Greenwich

Recognising moving face composites Recognising moving face composites  Karen Lander, Emma Portch & Charlie Frowd University of Manchester, Bournemouth University, University of Central Lancashire A consistently small, but robust, recognition advantage has been found for dynamic vs. static familiar faces (Lander, Christie & Bruce, 1999). There may be a generalised benefit for viewing a face moving naturally and / or each known face may have an associated characteristic motion signature, which acts as an additional cue to identity (O’Toole et al., 2002).  In a criminal investigation, facial composites are images constructed by witnesses and victims of people they have seen to commit crime. Under some circumstances, animation techniques have been found to improve composite recognition by people familiar with the target identities. In Study 1, composites of famous faces are animated in a number of different ways.  Our results suggest that the greatest recognition advantage is found for faces animated using their own characteristic movement parameters.  There was also an advantage for any movement compared with a static presentation.  In Study 2, we further compared recognition of famous face composites from static and moving displays.  Specifically, we found a larger recognition advantage when famous faces are shown animated using their own characteristic motion parameters and by general motion (a short animated sequence depicting the face morphing sequentially through caricature and anti-caricature phases, generated by EvoFIT), compared with own motion shown twice, general motion shown twice or a static presentation.  We suggest that motion may aid recognition in different ways and that these are additive in aiding recognition.  Results are discussed within both a theoretical and forensic / applied setting. 

Dr. Karen Lander
Senior Lecturer in Experimental Psychology, University of Manchester

Five minutes break


Historical Ethnographic Craniofacial Collections: incarcerated flesh with untold histories and modern dilemmas. Housed in the School of Anatomical Sciences at the University of the Witwatersrand is the “Raymond A. Dart Collection of African Life and Death Masks”. This comprises 1110 typology masks which have an uneasy association with early-twentieth century anthropology and Eurocentric imperialism, fascism, racism and eugenics. Related to this collection is a single full body cast of a young deceased woman, identified via historical and comparative craniofacial analysis to be a ‡Khomani San woman named /Keri-/Keri. In life, her face was cast during the July 1936 Wits Kalahari Expedition. From September 1936 till January 1937, she formed part of a “Bushmen Camp” attraction at the Empire Exhibition in Milner Park (Johannesburg). By 1939, her lone premature death in a colonial hospital led to further anthropological exploitation. Her full body was cast and dissected, with skeleton retained. Her cast and articulated skeleton were subsequently displayed as a macabre diorama. /Keri-/Keri’s case is not isolated. Numerous Africans were subjected to full body and facial casting for typological investigations, and to immortalise the declining Khoisan ‘race’ before presumed extinction. The mid-twentieth century eventually witnessed a withdrawal from typological concepts to those recognising the complex influences of genetics, environment and diet in biological variation. Though anthropology advanced, remnants of an undesirable past remain hidden in archives with uncertainties as to their appropriate storage, use and display. This presentation aims to generate an academic dialogue regarding the ethical function and treatment of such sensitive collections, while considering the social history and impact on affected nations.

Dr Tobias Houlton
Lecturer in Forensic Art and Facial Imaging Centre for Anatomy and Human Identification, University of Dundee (UK)

12:15 - 12:25

Human skull from Covo Valley. Traces of socio-religious practice. It was identified, in the reserves of the National Museum of Archeology (M.N.A.), with a view to preparing a study on trepanation in the Prehistory of Portugal, a human skullcap, coming from Pragança, offering ample traces of cuts. A first interpretation of such testimonies suggested the presence of extensive unfinished polygonal trepanation. However, closer observation and parallels with other occurrences allowed us to raise new hypotheses, suggesting the traces of cuts before the attempt to obtain a cranial mask, an aspect known, but uncommon, in Prehistory. Since its inclusion in the M.N.A. collections, that anthropological specimen it was found deposited on a natural "shelf", on one of the walls of the karst cavity, being accompanied by a flint knife, a bone drill and a piece of ceramic fragment. being accompanied by a flint knife, a bone drill and a piece of ceramic fragment. Little is known about the provenance of the skull now disclosed, discovered in a small underground cavity in the Pragança area, in the Serra de Montejunto, as part of work carried out there in the 1940s, under the direction of Manuel Heleno. That professor of Prehistory at the Faculty of Arts of Lisbon and former director of the M.N.A., by inherent position, succinctly references the finding of the aforementioned skull, in one of his many “field notebooks”, as was his habit. An anthropological analysis was carried out in order to try to obtain an interpretation for this specific case. Key words: Trepanation, Chalcolithic rituals, Bone mask, human skull. Mário Varela Gomes Carlos Didelet Archaeologyand Paleosciences Institute Nova Universtty of Lisbon cdidelet@gmail.com

Carlos Didelet
Archaeology and Paleosciences Institute – Nova Universtty of Lisbon

11:00 AM

Sensory Break


Lunch: Meet and Greet via zoom meeting. Please navigate to the networking lounge


On Seeing and Not Seeing Race This presentation offers a wider context for discussions of craniofacial identification by considering the troubled role of race as an object of expertise in forensic genetics and other emerging policing technologies. A range of developments have, in combination, increased the visibility and valorisation of supposedly obviously apparent racial differences. New technologies and techniques seek to rationalise identifications through and across racial differences but this goal remain elusive. Moreover, in settings where minorities already experience hyper-surveillance, descriptions and representations rooted in race reproduce group stigmatization and over-policing.

Prof. David Skinner
Associate Professor of Sociology at Anglia Ruskin University, Cambridge

Matching 3D facial shape to DNA-related properties Face recognition is a widely accepted biometric verification tool, as the face contains a lot of information about the identity of a person. In this study, a 2-step neural-based pipeline is presented for matching 3D facial shape to multiple DNA-related properties (sex, age, BMI and genomic background). The first step consists of a triplet loss-based metric learner that compresses facial shape into a lower dimensional embedding while preserving information about the property of interest. Most studies in the field of metric learning have only focused on 2D Euclidean data. In this work, geometric deep learning is employed to learn directly from 3D facial meshes. To this end, spiral convolutions are used along with a novel mesh-sampling scheme that retains uniformly sampled 3D points at different levels of resolution. The second step is a multi-biometric fusion by a fully connected neural network. The network takes an ensemble of embeddings and property labels as input and returns genuine and imposter scores. Since embeddings are accepted as an input, there is no need to train classifiers for the different properties and available data can be used more efficiently. Results obtained by a to-fold cross-validation for biometric verification show that combining multiple properties leads to stronger biometric systems. Furthermore, the proposed neural-based pipeline outperforms a linear baseline, which consists of principal component analysis, followed by classification with linear support vector machines and a Naïve Bayes-based score-fuser. S. S. Mahdi et al., 2021. 3D Facial Matching by Spiral Convolutional Metric Learning and a Biometric Fusion-Net of Demographic Properties," 2020 25th International Conference on Pattern Recognition (ICPR), pp. 1757-1764, doi: 10.1109/ICPR48806.2021.9412166.

Soha Sadat Mahdi and Nele Nauwelaers
Medical Imaging Research Center of KU Leuven
13:40- 13:50

Towards a fully automatic approach to craniofacial superimposition Towards a fully automatic approach to craniofacial superimposition Oscar Ibáñez a,b,*, Andrea Valsecchia b, Enrique Bermejob c, Rubén Martosd a Panacea Cooperative Research S. Coop., Ponferrada, Spain. bAndalusian Research Institute in Data Science and Computational Intelligence, University of Granada, Granada, Spain. c Departmentof Computer Science and Artificial Intelligence, University of Granada, Spain. dDepartment of Physical Anthropology, University of Granada, Granada, Spain. *Corresponding author:oscar.ibanez@panacea-coop.com Craniofacial superimposition generates controversy in the scientific community. It is a challenging, time-consuming and subjective comparison method where two different objects are compared, a facial photograph and skull (2D or 3D) image. Although this technique has been in use for a century, its great potentiality (it only requires facial photographs as antemortem data)has not positioned it as a commonly employed identification methods due to the lack of objective, reproducible, accurate and automatic approaches. Additionally, it is currently not possible to make firm statements about its overall reliability because studies of its reliability have used small samples and have not been replicated. In this presentation we will briefly introduce and analyse the last achievements obtained by researchers from Panacea Cooperative Research and the University of Granada which position the technique as a fully automated process with the following stages: cephalometric landmark location in facial photographs, craniometric landmark location in 3D skull models, pose and subject to camera distance estimation in facial photographs, soft tissue depth estimation from craniometric landmarks, skull-face overlay over one or multiple-photos at the same time, ranking of candidates and decision making. Keywords: Forensic Anthropology, Craniofacial Identification, Craniofacial Superimposition, Skull-face overlay, craniometric landmarks, cephalometric landmarks, pose estimation, subject-to-camera distance estimation

Dr. Oscar Ibáñez
Panacea Cooperative Research S. Coop., Ponferrada, Spain

Ten minutes break


From an other-race effect for algorithms to measuring bias in deep-learning based face algorithms In 2011, Phillips, Jiang, Narvekar, and O’Toole showed an other-race effect for face recognition algorithms. The study reported results for a Western algorithm made by fusing eight algorithms from Western countries and an East Asian algorithm made by fusing five algorithms from East Asian countries. At the low false accept rates required for most security applications, the Western algorithm recognized Caucasian faces more accurately than East Asian faces and the East Asian algorithm recognized East Asian faces more accurately than Caucasian faces. In a follow-up study, Cavazos, Phillips, Castillo, and O’Toole (2020), examined bias in deep-learning algorithms. They presented three findings. First, dataset difficulty affected both overall recognition accuracy and race bias, such that race bias increased with item difficulty. Second, for all four algorithms, the degree of bias varied depending on the identification decision threshold. To achieve equal false accept rates (FARs), East Asian faces required higher identification thresholds than Caucasian faces, for all algorithms. Third, demographic constraints on the formulation of the distributions used in the test, impacted estimates of algorithm accuracy.

Dr. Jonathon Phillips
National Institute of Standards and Technology's Information Technology Laboratory

Craniofacial analysis for Migrant Disaster Victim Identification This migrant disaster victim identification presentation is based on an 18-month British Academy funded project, which focused on the Canary Islands, clarifying the state of play of documentation and connections with West Africa: primarily with Senegal, which is described as the main origin of the migrants to the Canary Islands. With the collaboration of Italian and Spanish academics and the utilisation of Canarian data, we interrogate the challenges associated with the identification of migrant victims off the coast of the Canary Islands through fostered networks in the Canary Islands and Senegal. Finally, we present craniofacial depiction/analysis as an alternative biological and biometric tool for Migrant Disaster Victim Identification (MDVI). The presentation ends with a summary of the current status and provides recommendations for future MDVI.

Dr. Maria Castaneyra-Ruiz And Prof. Caroline Wilkinson
Face Lab, Liverpool John Moores University

Morphometric studies on the inferior nose in the context of forensic facial reconstruction using a German population sample Morphometric studies on the inferior nose in the context of forensic facial reconstruction using a German population sample Kevin Neuwirtha*, Ozgur Bulutb, Nicolle Freudensteinc, Katerina Harvatib aInstitute for Prehistory, Early History and Medieval Archaeology, Eberhard Karls Universität Tübingen, Germany b Paleoanthropology, Senckenberg Center for Human Evolution and Paleoecology, Eberhard Karls Universität Tübingen, Germany. c Freelance Forensic Anthropologist, Kaiserslautern, Germany. * Corresponding author: Kevin Neuwirth, Email:kevin.neuwirth@student.uni-tuebingen.de Abstract Due to its positioning in the center of the human face, the nose is a distinguishing feature of high importance. For this reason, special attention is paid to the modeling of its morphology during reconstruction of faces based on the cranial bones of deceased persons. Until now, however, extensive anthropometric studies of the soft tissue of the nose for an average German population, associated with measurements of its bony counterpart, have been lacking - this deficiency is addressed by this study.The inferior portion of the noses of 336 individuals (of whom 170 are male and 166 female) was measured using standardized landmarks, based on CT scans. In addition, morphological criteria were used to gain further insight regarding nasal morphologies and their relationship to metric data within the sample. The width of the apertura piriformis, the width of the nose, the width of the columella, the width of the nostrils, the distance from the most lateral and medial extent of the nostril to the apertura piriformis, and the distances between the most medial and lateral points of the nostrils to each other were calculated.Highly significant regression models could be established for both the calculation of the width of the nose and the calculation of the width of the columella based on the width of the apertura piriformis, which for the population concerned proved to be more efficient than conventional methods. In addition, alternative approaches for application in facial reconstruction could be developed. Sexual dimorphisms in varying degrees were found for the majority of the analyzed measurements and morphologies. The results of this study enable a more realistic reconstruction of the nose within a German population and are thereby of importance beyond the field of forensic facial reconstruction. Thus, a deeper understanding of nasal morphology is vital for medical applications such as plastic surgery or during the planning of surgical procedures. These disciplines equally benefit from the availability of local anthropometric databases. Keywords: Forensic anthropology, craniofacial reconstruction, anthropometric study

Kevin Neuwirth
Institute for Prehistory, Early History and Medieval Archaeology, Eberhard Karls Universität Tübingen, Germany

Five minutes break


Panel with keynote speakers: Prof. Amade M’charek, Prof. Josh Davis, Dr. Tobias Houlton, Prof. David Skinner, Dr. Jonathon Phillips Chair: Prof. Caroline Wilkinson

Supported by: