Wintics video analytics help SNCF spot people on the tracks

France's evaluation committee found that Wintics video analytics let SNCF detect people on the tracks more regularly. SNCF counted 660 true and 233 false intrusion alerts.

  • 2 an organization

    SNCF received 660 true intrusion alerts against 233 false ones, and the committee credits the tool with detecting people on the tracks more regularly.

  • Not rated society

    The committee names risks to people and equipment that the tool prevents but counts only three interventions owed to it during the Olympics.

Depth runs from 1, a convenience or time saved, to 5, a life or a livelihood preserved.

Status
Reported and intended benefits.
Evidence

foundstudiedreportedclaimed

An independent inquiry report supports the benefit.

Compared with

Video operators watching the camera feeds without the algorithm. Assumed without a measurement.

The system’s part
Not isolated. The committee counts SNCF's alerts and does not compare detection with and without the tool. Video operators checked each alert and decided on action. The committee states that the vendor trained the models before use and that they did not keep learning.
Limits
Not yet checked.
Costs and harms
Not yet checked.
Later
Not yet checked.

Download data Cite Suggest a correction

Benefit type
Reduced Risk
System
Algorithmic video processing developed by Wintics
Made by
Wintics
Used by
SNCF (rail security services), at stations in the Paris region
Where
Paris region, France
Stage
In a pilot
Date
Kind
Aggregate result
Broad benefit
Monitoring Processes, Materials, or Surroundings, Identifying Objects, Actions, and Events
Collections
Scrutiny and findings
Last revised
31 October 2026

Sources

  1. Inquiry report, by an evaluator that the makers or users commissioned.

    Comité d'évaluation de l'expérimentation de traitements algorithmiques d'images (chair Christian Vigouroux). Expérimentation de traitements algorithmiques d'images légalement collectées au moyen de systèmes de vidéoprotection, rapport du comité d'évaluation. Ministère de l'Intérieur, January 2025. § 3.2.1, p. 68, and § 2.3.5, p. 59.

    Il en est ainsi, en particulier, pour la SNCF et la RATP, à qui il permet de détecter de façon plus régulière la présence de personnes sur les voies, en prévenant ainsi des risques pour la sécurité des personnes ou du matériel (en cas de tentative de sabotage) et, accessoirement, en limitant les perturbations causées à l’exploitation du réseau.

  2. Inquiry report, by authors with no tie to the system.

    Sénat, commission des lois. Rapport d'information n° 374 (2024-2025) sur le bilan de la mise en œuvre de la loi du 19 mai 2023 relative aux jeux Olympiques et Paralympiques de 2024 (vidéoprotection intelligente et sécurité privée), by Françoise Dumont and Marie-Pierre de La Gontrie. Registered 19 February 2025.

    De même, les signalements relatifs à des intrusions représentent 57 % des signalements totaux opérés par le dispositif expérimenté par la SNCF (1 552 signalements), avec un taux de « vrais positifs » de 74 %.

Details

AI system

Workflow

The vendor trained the models before deployment, and they did not keep learning in use. Operators set the parameters for each camera during a calibration phase. Video operators checked each alert before any intervention.

Follow-up

Conditions for replication

The camera's view is well lit. Staff can reach the spot quickly after an alert. The station has too few agents to watch every zone.

Entry details

Sector

Railway safety / trespass detection.

Purpose

Wintics built its algorithmic video processing to alert operators to predetermined events in camera images, such as people entering forbidden zones.

Not yet checked: model, release, provider, role, configuration, inputs and outputs, autonomy, human oversight, AI contribution, beneficiaries, reach, distribution and access, duration, every measurement question, every money question, ownership and control, dependencies, cross-border connections, geopolitical implications, replication attempts, availability, institutional decision, grounds, action and result, success criterion, every evidence question, observed or intended, alternative assessments, related entries, other identifiers, category mappings.

Cite this record

Peter H. Vartanian. (2026). Wintics video analytics help SNCF spot people on the tracks (AIBD-0837). Artificial Intelligence Benefits Database, edition of 31 October 2026. https://petervartanian.xyz/aibd/records/AIBD-0837.html (archived at https://doi.org/10.7910/DVN/AADSII, version 4.0)

BibTeX Data Field definitions