PRESS RELEASE - TAIL is a diagnostic tool that automatically detects when a pig bites the tail of another pig.
This innovation combines a removable video camera, placed in a collective box of post-weaning and fattening pigs, and an artificial intelligence and computer vision cloud application that autonomously detects tail-biting behaviour.
This tool is a major advantage in terms of behavioural observation, as it detects tail biting in a non-intrusive manner instead of the veterinarian or technician. The events searched for are automatically detected by the AI algorithms and returned to the user in the form of statistics.
The analysis is at least 3 times faster than an analysis performed by a human. Its performance remains constant, as the algorithms do not experience fatigue or loss of attention, making the diagnosis more objective.
In the future, knowledge of these data will make it possible to measure the behaviour of pigs and then to establish prediction models to anticipate the biting phenomenon.
The creation of TAIL and its originality result from a multidisciplinary development approach combining zootechnical expertise and artificial intelligence.
The development of our solution for the detection of caudophagy required the use of 2 technologies:
For this project, the Dilepix team set up a new type of neural network allowing spatio-temporal analysis.
Until now, artificial intelligence analysis of videos has been carried out frame by frame, resulting in a large number of false positives for behaviour that lasts over time.
These new neural networks now take into account the context before and after an event.
A technological solution like TAIL will help the industry to support changes in practice:
This tool does not require the installation of any sensors on the animal.
The camera monitors from a distance and does not disturb the animals' rhythm. It has no influence on their behaviour.
Caudophagy is the result of several factors (boredom, unsuitable temperature, disturbing activity, feeding problems, etc.). Understanding its origins will eventually make it possible to establish predictive models to anticipate this phenomenon.
When the analysis of animal behaviour is done manually, it means hours of video for technicians and veterinarians to watch.
Although necessary, this approach to observing animals is not widely used, as it involves many constraints:
Once the videos have been transmitted to the cloud analysis platform, the events searched for are automatically detected by the artificial intelligence algorithms and returned to the user in the form of statistics.
A solution 3 times faster than a human
A veterinarian would take on average 3 hours to analyse 1 hour of video. On the Dilepix Cloud, the analysis takes less than an hour and only 5 minutes for the veterinarian (corresponding to the file deposit time), because he can continue his activities while the platform analyses the video for him.
A camera-based solution observes natural animal behaviour that may be interrupted or altered by a human presence, especially a technician or veterinarian who is not a permanent staff member of the farm. Our solution provides better ground truth than human observation.
Only three manipulations are required:
Dilepix will take advantage of its presence at the next SPACE de Rennes (13-15 Sept 2022) to present its automatic livestock detection solutions.
đŸ‘‰We look forward to seeing you in HALL 4 - Stand Agretic (B43)
Of course, the main activity will be livestock farming, but we also have topics in Machinery & robotics and industrial insect production.
đŸ‘‰ Highlights
Presentation of "TAIL", the software for automatic detection of tail-biting in pig farms, which combines camera vision and artificial intelligence...
| Pig production | Cattle production | Insects | Robotics & machinery |
| Software for automatic detection of sow ovulation | Software for measuring the activity of dairy and suckler cows | LSoftware for automatic counting of Black Soldier Fly larvae and mealworms | Software for camera control of robots and agricultural machinery |
| Software for measuring pig activity | |||
| Software for measuring the weight of fattening pigs | |||
| Caudophagy detection sofware |
About Dilepix
Dilepix's know-how combines expertise in artificial intelligence (neural networks developed in-house by its engineers) with advanced skills in image analysis.
The combination of these two technological building blocks enables ultra-precise detection and localisation of agronomic threats and opportunities.
Developed in laboratory and proven by major manufacturers, the technology used by Dilepix is patented.
Founded by Aurélien Yol, a former research engineer at Inria and Alban Pobla, a serial entrepreneur, Dilepix was created in 2018.
Hosted and supported by Inria's Startup Studio and then by Le Village by CA Ille-et-Vilaine until 2020, the young company is now flying under its own steam and has 14 employees.
Dilepix was a winner of the prestigious iLab competition [1] in 2019 and participated two years in a row (2018 and 2019) at CES in Las Vegas.
It raised a first seed round of €1.7 million in 2018 and has just completed an intermediate financing round. In the medium term, Dilepix wants to partner with a manufacturer to ensure its development and strengthen its product roadmap.
Dilepix addresses producer groups, veterinary groups and pharmaceutical and consulting companies that are looking to integrate more innovation into their product and service offerings.
Dilepix is derived from the scientific name of the chameleon "dilepis", an animal with excellent visual skills.
[1] national competition for the creation of innovative technology companies created in 1999 by the Ministry of Higher Education and Research http://www.enseignementsup-recherche.gouv.fr/pid24633/concours-d-innovation-i-lab.html
Contact : aurelie.busnel@dilepix.com