HOME / CASES / Pig farming

Non-invasive piglet weighing: a unique solution for agribusiness companies

Technology for pig breeding companies to improve the efficiency of pig fattening through daily remote weighing of pigs by applying neural network video analytics CVC
One of the keys to successful piglet fattening is having accurate data on piglet weight changes. Without this, it is extremely difficult to optimize the fattening strategy and to predict the slaughter weight of pigs at finishing. But until now, pig producers have not had an effective, yet affordable, tool for obtaining this information. We are proud to tell you about the solution we are successfully implementing at Dymov Group of Companies.
Contents list

The essence of the problem and how it has been solved until now

The main objective of fattening is to maximize the weight gain of the piglet with minimal feed inputs. Of course, in order to influence the process, it must be controlled. In this case, the first place to go is to regularly determine the weight gain of individuals. These indicators will allow you to timely change the composition of feed additives in the direction of its efficiency. By the way, the cost of such additives can reach up to 30% of the cost of the feed base.

In order to evaluate the effectiveness of additives and adjust them, swine complexes need to receive information on the exact weight gain of the average machine at least every 2-3 days.

A person who is not involved in the industry will say, "You could just weigh the piglets at regular intervals. However, pig farmers realize that this solution is unrealistic for a number of reasons.

Firstly, piglet anxiety reduces fattening efficiency many times over. During stress, piglets lose their appetite, the quality of their meat decreases, and in some cases the animal may even die.

Secondly, from the point of view of epidemiological danger during the fattening period it is necessary to minimize the contact of piglets with people.

And, thirdly, in one hangar there are about 40-50 pens, and each pen contains from 35 to 120 piglets. If the company has 1100 such machines, it means that there can be about 60,000 piglets in the fattening process at the same time. Thus, it is necessary not only to create a whole infrastructure for weighing, but also to create a separate staff for this purpose. To put it another way, weighing piglets manually will not work.

At the moment, the pig industry is getting out of this situation by selective weighing. In the selective inspection process, a few individuals are selected and the weighing results are spread over the entire pen. This method is similar in efficiency to statistical population surveys, where the margin of error with the actual data can be significant. In other words, when the piglets turn into boars, the pig farm management may be very surprised by the real picture. For example, instead of the planned 120 kilograms per pig, only 80 kilograms will be produced. With this underweight, the company will lose the value of several tons of pork.
The pork industry has lived in this situation for decades. But in the age of neural network technologies, many previously unsolvable problems have a cost-effective solution.

Disadvantages of existing solutions on the market

To begin with, many people have tried and are trying to solve this problem. Three months after the start of our pilot, we learned that one of the world leaders in innovations for agricultural complexes has already presented equipment and a solution to this problem. Having studied the solution of this company, we were sincerely pleased....

Firstly, we got another confirmation that this problem is at the level of the world scale and is relevant. Secondly, when we saw the cost of the proposed equipment (by the way, it is European), we realized that our solution is much cheaper. For comparison, the cost of equipment for one machine reaches up to 10,000 euros. Thus, in comparable units, our equipment is 10-15 times cheaper than its Western counterpart.

In addition, there is an important security issue - where the servers with data processing are located. Location of servers in Holland, Denmark, etc. in the current reality is not quite suitable for domestic companies. Moreover, many customers want solutions in which the information will not leave their I.T. loop, i.e. the server must be located at the customer's site.

Also, existing solutions have a number of limitations, making the equipment extremely vulnerable to sinks. Namely: stereo cameras are used, additional sensors that must be placed fixed at a height of 2.5 meters. I think specialists will immediately realize how inconvenient this is, because in the process of washing, dirt flies in all directions. In other words, before each washing this equipment must be removed, disconnected, and after washing one and the same to mount it back. Agree, not very convenient....

We are fine with a competitive race and the presence of several players. The main question is who will have a solution that is more efficient in terms of cost of ownership, more reliable in terms of operation in real conditions. And, most importantly, how accurately the weight and size model will work.

The essence of our solution for non-invasive piglet weighing

The beginning of the project and a few words about the custome

So, we were approached by the IT department of the Dymov Group of Companies. The company unites a number of large agricultural enterprises in Russia in various business areas, including pig farming. Having voiced our main pain, we were interested in solving this problem. Especially when we learned that this problem is a mass one for all companies in this industry, we immediately caught fire in front of a new non-standard and interesting task.

Yes, our company has a great experience in implementing various solutions based on artificial intelligence and computer vision for all kinds of tasks of companies from different spheres of activity. But this was the first time we encountered pig farming.

    How CVC has met the challeng

    The essence is as follows - cameras see a piglet and analyze its dimensions. The pig is not a static object, i.e. it can run, lie down, turn its head and so on. That is, we have before us objects of unequal geometry. Accordingly, the first question is to measure these objects, and the second question is to translate the indicators into weight. Thus, to create a weight-size model, which, given the given parameters (density, width, height, neck circumference, etc.) determines the weight of the object.

    One important consideration in non-invasive piglet weighing is that such a mass-dimensional model must account for weight differences between individual piglets. For example, at the same point in time, there may be piglets in the pen that are significantly different in weight (the largest piglet may weigh 80 kg and the smallest 60 kg).

    If you have studied our other cases, you may have noticed that we always share our technologies openly. However, our solution for non-invasive piglet weighing is truly unique, so it is a trade secret. So we will leave the details out of it this time.

      Integration features

      The project was launched in the summer of 2021 and has gone through several phases:
      1. Design;
      2. Pilot installation;
      3. Development of mass and dimensional model;
      4. Verification of the results.
      Our technology has already shown positive results in three pig fattening cycles. We would like to make it clear that we are not even talking about a separate project, but about developing the technology from scratch. Many of the shortcomings of existing solutions have been taken into account in the integration process.
        • 1
          Equipment security. The first point was to ensure the survivability of the equipment in real conditions. As we said earlier, we are talking about the washing process, when dirt is uncontrollably flying in all directions. The cameras were placed under the ceiling much higher than 2.5 meters. As a result, equipment maintenance was reduced to annual wiping of the cameras.
        • 2
          Servers at the customer's site. For testing, we used our own server, specially built for the project. At the same time, our solution provides for the fact that during scaling the servers will be located at the customer's location. in this way, the issue of limited Internet access at such sites is completely solved.
        • 3
          Reasonable cost. Our solution is based on SaaS. At the same time, the cost of the subscription fee is 2-2.5 times lower and it is denominated in the national currency.
        • 4
          Report availability. Reports on weighing results are available at any time. The current system is set up in such a way that reports are generated once every 2-3 days. But it is possible to view the previous day's growth in visual graphs, which are easy to analyze, every day.
        Weighing of piglets by CVC neural network analytics takes place 12000 - 16000 times a day. At the same time, the system has no effect on the piglets themselves - they live a normal life.

        Interface illustration

        How quickly we set up the system at your site

        Since the model of non-invasive piglet weighing is now developed, we only need a few weeks from the time of placing the equipment to customize the system for your specific agribusiness. Contact us and in a couple of weeks it will be up and running.

          On the results achieved and the accuracy of measurements

          Of course, the first thing that is important to customers is whether the system gives accurate results on piglet weight. The only way to find out the answer to this question is to weigh the piglets after the end of the cycle and compare the data with the system results.

          The customer in this case study has been fattening piglets for 4 months. So far, we have successfully completed three cycles - three times for four months. After each cycle, each piglet is weighed and the actual results are determined. So the weighing results of our model differ from the actual results in 9 out of 10 cases by less than 3%, and in 1 out of 10 cases by 7%
            CVC neural network-based video analytics is your trusted expert in the pork industry, capable of monitoring weight gain in each of hundreds of fattening machines around the clock.

            When to start?

            Pig breeding is a rather competitive industry. Using this solution is a real opportunity for any company to reduce the cost of production and gain a competitive advantage from the introduction of artificial intelligence.

            The solution for pig farming
            Our technology provides a unique opportunity to analyze the weight gain of livestock, optimize feed mixtures and feeding programs, accurately determine feed conversion. And to do all this without stress for the animals and without human involvement.
            If this is your industry and your case, the question is not whether to use artificial intelligence or not. The question is when you do it: before your competitors or after. Everyone will have to switch to such a weighting system, the only question is who will get the advantage earlier: you or your competitors.

            We would also like to add that every successful project of this kind is a consequence of our company's ideology - the creation of industrial standards for neural network analytics. Creating an industrial standard implies its availability in terms of speed and resources, including money. For example, this technology makes it possible to use simple and inexpensive video cameras in a pig farm.
            Letter of thanks from Dymov Group of Companies
            Неинвазивное взвешивание свиней отзыв
            "...Dymov Company is interested in the development of the technology and its industrial application, and believes that the Power Aicontrol LLC team is an expert in creating advanced technologies based on machine vision and artificial intelligence".

            A. Khairetdinov
            Executive Director
            Dymov Group of Companies
            Cyber Vision Control
            CVC mitigates losses in any business, however small or big it is, whether it is a manufacturing enterprise
            or a service company.

            20 days
            Average pay-off period after CVC has been implemented
            >10 mln
            >100 mln
            Disciplinary violations detected
            Economic violations detected
            Read related
            Industry Solutions | Cases | Blog | About | Contact Us

            +7 (800) 301-39-70 contact@cvc.ai

            3, ul. Mira, ROOM 1H, St. Petersburg, 197101, Russia

            OGRN (company number) 1197847127125

            Privacy policy