FUSION – Film Understanding through Spectroscopy and Intelligent Optical Notch recognition
How can we improve the identification of old photographic films to better prevent their deterioration?
The KIK-IRPA Photo Library documents all aspects of Belgium’s cultural heritage through photographs dating from the late 19th century to the present day. Belgium thus has one of the most extensive photographic coverages of its heritage. The collection is also unique in that it preserves the original negatives for each image, unlike most photo libraries, which mainly consist of paper prints or slides.
Problem statement
Effective preservation of photographic collections requires precise knowledge of the chemical nature of film supports (nitrate, acetate, or polyester). This identification is essential for anticipating degradation processes, defining conservation priorities and treatments, and assessing health (for staff and collections) and safety risks (such as fire hazards).
Polyester is considered chemically stable, whereas cellulose nitrate and cellulose acetate are inherently unstable polymers. Over time, and particularly under unsuitable storage conditions, these materials degrade, releasing nitrogen oxides in the case of nitrate films and acetic acid in the case of acetate films (the so-called “vinegar syndrome”).
These degradation processes are autocatalytic. A chemical reaction is autocatalytic if one of the reaction products is also a catalyst for the same reaction. So after an initial period, the acids released from the films accelerate further deterioration and can even spread to nearby films.
While early photographic films were made exclusively from cellulose nitrate, its high flammability led manufacturers to develop alternative flexible film supports. Since the early 1920s, multiple types of film have coexisted on the market, meaning that dating alone is no longer sufficient to determine their composition. Today, identification relies on a combination of dating, visual inspection, and sometimes chemical testing—an approach that remains time-consuming and not always precise.
Perforation codes (“notch codes”) and edge markings (“edge printing”) applied by manufacturers may provide useful clues. However, their interpretation is complicated by the lack of a comprehensive reference system, the reuse of codes, and occasional duplication of markings. Although commonly used in identification methods, recent research has called into question the reliability of notch codes as a dating tool.[1]
This raises the following question:
How reliable are notch codes as a primary criterion for identifying photographic film?
[1] Garratt, C. (2025). Finding Safety in Notches: Using Atr-Ftir Spectroscopy and Industry Literature to Identify the Substrates of Photographic Negatives. AICCM Bulletin, 1–12. https://doi.org/10.1080/10344233.2024.2445869
Methodology
The project focuses on developing an AI-based tool capable of automatically detecting and extracting notch codes from digital images of negatives, and comparing them with manufacturers’ reference data. This work is based on a corpus of approximately 22,000 sheet films preserved at KIK-IRPA.
The results will be complemented by archival research (manufacturers’ records and KIK-IRPA archives) to confirm the identification of the different film models.
In a second phase, a non-invasive chemical analysis of the films will be carried out using Fourier-transform infrared spectroscopy (FTIR), on representative samples of each identified type.
By combining these datasets, the project will assess whether a reliable relationship exists between notch codes and the chemical composition of the film supports, and evaluate how useful these codes are as an identification tool.
Objectives of FUSION
- To develop an artificial intelligence tool capable of automatically identifying and classifying photographic films based on their digital reproductions.
- To establish a reliable database of film types held in the KIK-IRPA’s collections, linked to their chemical composition.
- To assess the reliability of notch codes as an identification criterion, whether used alone or in combination with other methods.
Expected impact
Providing a clear, scientifically grounded answer on the actual reliability of notch codes will bring valuable clarity to professionals managing photographic collections.
Regardless of the outcome, the development of an innovative system for the automatic recognition and classification of film models based on digital reproductions will make it possible to gain insights into large film collections more quickly. Although the system is developed using the KIK-IRPA collection and primarily intended for its use, it will be designed as a generic tool that can be applied in other institutions as well.
By combining expertise in documentation, conservation, laboratory analysis, and artificial intelligence, FUSION will also enable KIK-IRPA to further strengthen its strategy for managing this collection which is an invaluable documentary resource for Belgium’s cultural heritage.
Want to find out more? Get in touch.
KIK-IRPA collaborators
External contributors:
Joris Pockelé (Master’s student in Computer Science Engineering at Ghent University) – Research and development of the AI algorithm, in collaboration with Nicolas Nadisic
Clara Giunta (intern at the MatCoRe Unit at KIK-IRPA) – Chemical analysis of films using Fourier-transform infrared spectroscopy (FTIR), in collaboration with Francisco Mederos-Henry and Amandine Colignon
EEHOP – Electromagnetic Exploration of Historical Oil Paintings
The EEHOP project (Electromagnetic Exploration of Historical Oil Paintings), a collaboration between the Royal Institute for Cultural Heritage (KIK-IRPA) and UCLouvain, funded by the F.R.S.-FNRS, explores an innovative approach: using electromagnetic characterisation techniques to study the hidden transformations in oil paints.
