1. Introduction: The Challenge of Livestock Counting and the Promise of Artificial Intelligence and Drones.

Accurate livestock counting is a fundamental task for the efficient management of livestock farms. It allows for detailed inventory control, facilitates breeding and feeding planning, contributes to the early detection of health problems, and is essential for fair and transparent commercial transactions. However, traditional methods employed for this task are inherently laborious and susceptible to human error. This reality imposes significant limitations on the ability of producers to obtain reliable and up-to-date data on their herds.

In this context, the emergence of new technologies offers a promising horizon. Unmanned aerial vehicles, commonly known as drones, have become established as versatile and increasingly accessible aerial platforms for capturing visual information in agriculture. Equipped with high-resolution cameras, these devices can overfly large areas of land quickly and efficiently, obtaining detailed images and videos of livestock. In parallel, artificial intelligence (AI) has experienced exponential advances in the field of image and video analysis. Through the use of sophisticated algorithms, AI can process large volumes of visual data automatically, identifying patterns, detecting objects, and performing counting tasks with increasing accuracy.

The strategic combination of drones and artificial intelligence emerges as a solution with the potential to radically transform the livestock counting process. This technological synergy promises to overcome the limitations of traditional methods, offering producers tools capable of providing efficiency, accuracy, and new perspectives on the management of their animals. This report aims to analyze in depth how to address the problem of livestock counting through the use of images or videos captured by drones and processed with artificial intelligence, detailing the technology involved and the necessary requirements for its implementation, with a specific focus on the context of the Buenos Aires region, Argentina.

2. The Legacy of Counting: Traditional Methods and Their Limitations.

Throughout history, various techniques have been used to carry out livestock counting. One of the most primitive ways consisted of using physical objects, such as "tally sticks" or notches in wood, where each element represented an individual animal [1]. This method, based on one-to-one correspondence, was useful for small herds, but its practicality was severely compromised as the number of animals increased. The absence of a digital record also made historical tracking and long-term management difficult.

With the development of livestock farming, techniques such as grouping animals in corrals were adopted to facilitate visual counting or estimation of herd size [2]. However, the accuracy of these methods decreases significantly with the increase in herd size and animal density. The subjective nature of visual estimation introduces a considerable margin of error, influenced by the observer's experience and environmental conditions.

In more extensive operations, manual aerial censuses were used, carried out from light aircraft or helicopters, where human observers counted the animals from the air [4]. Although this technique allowed large areas to be covered, its high operating cost, the need for specialized crew, and the susceptibility to counting biases and observer fatigue represented significant limitations. The speed and altitude of flight could make the identification and accurate counting of each animal difficult.

Traditional methods of livestock counting share a series of inherent limitations and challenges. Manual counting is intrinsically tedious, requires a large investment of labor and time, which often leads to counts being carried out irregularly [5]. The allocation of human resources for this task diverts attention from other crucial activities in livestock management, and the lack of frequent data prevents constant and effective monitoring. In addition, fatigue, lack of concentration, and adverse environmental conditions can lead to significant errors in the count [5]. Even small error rates can translate into substantial economic losses in large herds [8].

Access and counting in extensive pastures or complex terrains present considerable logistical challenges [4]. Ground methods can be impossible or extremely inefficient in certain geographies, where the topography of the land hinders access to all areas of the pasture and the dispersion of livestock over large areas prevents a comprehensive count. On the other hand, methods that require animals to pass through narrow points to be counted can generate stress and be harmful to their welfare, especially in the case of pregnant females [6]. Finally, most traditional counts are carried out infrequently, which limits the ability of producers to detect problems such as theft or diseases in early stages . This outdated information can lead to suboptimal management decisions, with significant financial and health consequences.

3. The AI Revolution: Intelligent Object Counting in the Visual World.

Artificial intelligence (AI) has emerged as a field of computer science dedicated to the development of systems capable of performing tasks that normally require human intelligence. In the context of counting objects in images and videos, AI combines the power of machine learning algorithms with advanced image processing techniques to identify and enumerate distinct elements within digital visual data [10]. This technology can differentiate between various types of objects, sizes, and shapes, even in crowded or dynamically changing scenes [12].

The key components of AI for object counting include detection, tracking, and counting [10]. Detection is the first step, where the AI system identifies individual objects in the image [12]. Using deep learning models, such as convolutional neural networks (CNNs), the system can recognize and locate various objects with high accuracy [12]. In video sequences or crowded scenes, tracking algorithms maintain the identity of each object as it moves or interacts with other elements, ensuring accurate counting over time and space [12]. The final step adds the detected objects, providing a total count [12]. Advanced AI models can count objects even in dense and overlapping scenes, where traditional counting methods fail [12].

Within the field of AI, deep learning has become the predominant technique for visual analysis. Convolutional neural networks (CNNs) are a specific type of deep learning architecture that has demonstrated exceptional effectiveness in computer vision tasks, including object detection and counting [10]. CNNs are designed to process data with a grid-like structure, such as images, and have the ability to automatically learn distinctive features of objects directly from pixel data. This ability to automate feature extraction overcomes the limitations of traditional methods based on manually designed features. Instead of explicitly defining what to look for, CNNs learn complex patterns directly from the data, making them more robust to variability in the appearance of objects.

AI for object counting is a versatile technology that has found applications in various sectors [12]. In manufacturing, it is crucial for ensuring the accuracy of parts and products on assembly lines and for quality control [10]. In logistics and supply chain management, it optimizes operations by automating the tracking of goods [12]. Warehouses benefit from better inventory management [12]. In office environments, it contributes to resource management and space optimization [12]. In the retail sector, object counting systems powered by AI can monitor product levels on shelves, providing real-time inventory data [12]. It is also fundamental in traffic and crowd management [12], as well as in wildlife conservation and research, where it allows for estimating animal populations from aerial images or camera traps [10]. The wide range of successful applications in multiple industries suggests the significant potential of AI to address the challenge of livestock counting. The automation and accuracy that AI brings to these fields are directly transferable to the problem of livestock management.

4. Eyes in the Sky, Brain on the Ground: AI Techniques for Livestock Counting with Drones.

The application of artificial intelligence to livestock counting using aerial images captured by drones involves the use of specific computer vision techniques. One of the main ones is object detection, whose objective is to identify and locate the presence of individual animals in each frame of an image or video [11]. There are various object detection algorithms, each with its own strengths and weaknesses. YOLO (You Only Look Once) stands out for its speed, making it suitable for real-time processing [11]. SSD (Single Shot Multibox Detector) is also a fast and efficient algorithm [18]. Faster R-CNN (Regions with CNN features) usually offers greater accuracy, especially in the detection of small objects [18]. Mask R-CNN (Mask Regions with CNN features) goes a step further by segmenting the individual instances of each detected object, which can be useful for differentiating animals that are very close together [21]. CenterNet simplifies detection by focusing on identifying the centers of objects instead of predicting the boundaries of the boxes [18]. The choice of the most suitable algorithm will depend on the specific requirements of the livestock counting project, considering factors such as the necessary processing speed and the importance of accuracy in the detection of small or overlapping objects [22].

The use of drone images for livestock counting presents specific considerations. The scale of the animals can vary significantly depending on the flight altitude, and occlusion due to the proximity of the animals is a common challenge [17]. In addition, lighting conditions can change rapidly during a flight. Therefore, AI models must be robust to these variations to achieve accurate counting. Training the models with diverse datasets that reflect these conditions is crucial to ensure their performance in different scenarios. The training images should include livestock captured from various altitudes and angles, in different grouping densities, and under different lighting and weather conditions.

When using drone videos, object tracking becomes an essential technique to avoid double-counting moving animals [10]. Tracking algorithms assign unique identifiers to each detected animal and track them through the frames of the video, even when they are partially hidden or their appearance changes temporarily [11]. ByteTrack offers a balance between accuracy and speed, with lower computational complexity [13]. BoT-SORT provides greater tracking accuracy, especially in challenging scenarios with occlusions and camera movement, although at the cost of greater computational complexity [13]. Platforms like Ultralytics YOLO integrate tracking functionalities, facilitating the implementation of complete detection and counting solutions in video [13].

Despite the advances, livestock counting with AI from drone images presents specific challenges. One of them is the differentiation between animals of different species or even breeds within the same species [28]. AI models must be trained to recognize the distinctive visual characteristics of the specific type of livestock to be counted. Another challenge is to achieve accurate counting in crowded conditions or when animals overlap in the images [10]. Algorithms must be able to distinguish individuals even when they are very close to each other [30]. Robustness to environmental conditions, such as variable lighting, adverse weather, and image quality, is also fundamental [7]. Finally, it is important to avoid misclassifying other objects present in the rural environment, such as humans or dogs, as livestock [7]. To address these challenges, AI models must be trained specifically with data that includes examples of livestock and other common objects in pastures, allowing them to learn to discriminate based on relevant visual characteristics [34].

5. The Ideal Drone for the Modern Livestock Farmer: Key Selection and Specifications.

Choosing the right drone is a determining factor for the success of livestock counting using AI. There are mainly three types of drones relevant to this application: multirotors, fixed-wing, and hybrid (VTOL). Multirotor drones offer advantages such as vertical takeoff and landing, hovering capability, and good maneuverability, making them ideal for detailed inspections in specific areas and counts in smaller pastures . However, they usually have shorter flight times and are less resistant to wind . On the other hand, fixed-wing drones are characterized by greater autonomy, the ability to cover large areas, and greater flight efficiency, making them suitable for monitoring large pastures and counts in extensive areas of land . Their main disadvantage is the need for space for takeoff and landing, as well as lower maneuverability . Hybrid drones (VTOL) combine the advantages of both types, offering vertical takeoff and landing with the flight efficiency of fixed-wing drones, which provides them with great versatility, although they are usually more expensive and complex .

When selecting a drone for livestock counting, it is crucial to consider several key factors. Flight time must be sufficient to cover the desired area without the need for multiple recharges . Autonomy is directly related to battery size and drone efficiency . Greater autonomy reduces the total mission time and the need for additional batteries. Camera quality is fundamental for the identification and accurate counting of animals [35]. The resolution and type of sensor (RGB, thermal, multispectral) should be considered according to specific needs [38]. High resolution allows for the detection of animals even from high altitudes, while thermal cameras are essential for counting in low light conditions [37]. Flight stability is important for obtaining sharp images and videos, especially in windy conditions . Drones with advanced stabilization systems (gimbal) are preferable to minimize motion blur . The payload capacity of the drone will determine what types of cameras and additional sensors can be carried simultaneously . If the use of multiple cameras is required, the drone must have sufficient payload capacity. Resistance to environmental conditions (wind, rain, temperature) will increase the window of opportunity for carrying out counting flights . Finally, ease of use and the availability of intuitive flight planning software are important for optimizing area coverage and data collection [9].

There are various drone models on the market that may be suitable for livestock counting. Some examples include the DJI Mavic 3 series (in its Thermal and Multispectral versions), the DJI Matrice series, the JOUAV CW-30E, the Autel EVO Max 4T, and the ZenaDrone 1000 . The following table offers a comparison of some relevant models:

This table provides an overview of the available options and their key attributes, helping the user to evaluate the trade-offs between different models according to their needs and budget.

6. Seeing Through the Lens: Cameras and Sensors for Accurate Counting.

The choice of the right camera and sensors for the drone is essential to obtain quality images that allow for accurate counting of livestock using AI. Mainly, RGB and thermal cameras are used for this application , although multispectral cameras can also provide valuable information about the environment [12].

RGB cameras work by capturing images in the visible spectrum of light, similar to a conventional photographic camera . This provides high-resolution and color images that are easy to interpret visually and that allow for the identification of individual characteristics of the animals, such as their size, shape, and color . The high resolution of these cameras facilitates the detection of animals even from high altitudes . However, their performance is limited in low light conditions or at night , and they may have difficulty detecting animals that are camouflaged or under dense vegetation . The dependence on visible light restricts the effective operating hours.

Thermal cameras, on the other hand, detect the infrared radiation emitted by objects in the form of heat, which allows them to "see" thermal signatures . This has the great advantage of allowing the detection of animals even in total darkness , through light fog or vegetation . They are especially useful for conducting night censuses and for identifying animals with elevated body temperatures, which could indicate a health problem . The independence from visible light significantly expands the operating hours and the detection capability in various conditions. However, thermal cameras usually have a lower resolution compared to RGB cameras , which can make it difficult to identify visual details or distinguish between species based solely on the thermal signature . Weather and dense vegetation can also affect the accuracy of thermal detection .

Multispectral cameras capture images in multiple bands of the electromagnetic spectrum, including the visible and near-infrared [12]. While not the primary sensor for direct animal counting, they provide valuable information about vegetation health, which can be useful for pasture monitoring and the assessment of livestock nutritional condition.

The combination of RGB and thermal cameras on the same drone can offer a more complete solution for livestock counting . RGB images can be used for visual identification and counting during the day, while thermal images can complement the counting at night or in low visibility conditions, and help detect animals that are hidden under vegetation . The visual information from the RGB camera can even be used to classify the thermal signatures detected by the thermal camera, improving the overall accuracy of the system .

7. Unleashing the Power of Data: Software and AI Processing Platforms from Intermedia IT.

Once drone images or videos have been captured, it is necessary to use specialized software and platforms to process the data using artificial intelligence algorithms and obtain the livestock count. Intermedia IT offers advanced solutions to address this need, providing producers with the ability to perform accurate and efficient counts through a desktop product or a cloud platform.

Intermedia IT specializes in the development of customized solutions, adapting to the specific needs of each producer and the particular problems of each case. Whether a solution is required for a quick and specific count using desktop software, or a cloud platform for the automated processing of large volumes of data and continuous monitoring, Intermedia IT can develop the appropriate tool.

The AI-powered livestock counting solutions from Intermedia IT are based on deep learning algorithms and state-of-the-art computer vision techniques, ensuring high accuracy in animal detection and counting. The flexibility of its products allows producers to choose the modality that best suits their infrastructure and operational requirements.

By selecting the software solutions and AI processing platforms from Intermedia IT, producers can expect high accuracy in livestock counting algorithms, user-friendly and intuitive interfaces, the capacity to handle large volumes of drone data, and the possibility of customizing AI models according to their specific needs.

8. Navigating the Legal Framework: Regulations and Permits for Drones in Buenos Aires, Argentina.

The operation of drones for livestock counting purposes in the Buenos Aires region, Argentina, is subject to the regulations established by the National Civil Aviation Administration (ANAC) [47]. ANAC classifies drones according to their weight and use, whether recreational or commercial . Class A drones (up to 500 grams) are exempt from registration for recreational use, while Class B (500 grams - 5 kg), Class C (5 kg - 25 kg), Class D (25 kg - 150 kg), and Class E (more than 150 kg) require mandatory registration in the National Aircraft Registry of ANAC .

The requirements for registration include the operator being over 18 years of age (or 16-17 with adult supervision) , payment of the corresponding fees , submission of a declaration of ownership with a certified signature and, in some cases, a photograph of the drone's serial number and proof of CUIT/CUIL .

There are important operational restrictions that must be met. It is prohibited to fly within 5 km of airports, airfields, and heliports , over densely populated areas or crowds , at an altitude greater than 122 meters (43 meters in controlled airspace) , flying is only allowed during the day and in good weather conditions , always maintaining visual line of sight with the drone (VLOS) and not operating within 30 meters of buildings . It is also prohibited to fly in sensitive areas such as government or military facilities , and a minimum horizontal distance of 30 meters and a vertical distance of 10 meters from people must be maintained .

For commercial operations, such as livestock counting for profit, additional permits from ANAC are required . This generally involves obtaining authorization as a "remote crew member," presenting a psychophysiological fitness certificate, having civil liability insurance for possible damages to third parties, having an operations manual, and a risk management system . While general regulations apply to the use of drones in agriculture, there may be specific additional considerations for this activity . It is strongly recommended to consult the official ANAC website directly for the most up-to-date and detailed information on current regulations in the Buenos Aires region .

9. Feeding the Intelligence: Data Requirements for Accurate Counting.

Training artificial intelligence models for accurate livestock counting from drone images or videos requires a significant amount of high-quality data [51]. The amount of data is a crucial factor; generally, a large volume of images or videos that have been meticulously annotated is needed so that the model can learn to recognize and count livestock effectively [51]. The greater the variability in the training data, including different viewing angles, lighting conditions, weather conditions, and livestock breeds, the more robust and accurate the final model will be [51]. A model trained with limited data may have difficulty generalizing to new images or videos that do not resemble the data used during training.

The quality of the data is equally important. Images and videos must have high resolution and clarity to allow for accurate identification of the animals [9]. Motion blur or low resolution can hinder detection and counting, even for an advanced AI model. If the animals are not clearly visible in the images, the model will have difficulty learning to recognize them.

In addition to quantity and quality, data annotation is a fundamental step in the process of training AI models for object counting. Images and videos must be meticulously annotated, indicating the location of each animal present. This is usually done by creating "bounding boxes" (delimiting rectangles) around each animal or through segmentation techniques that precisely delineate the shape of each animal [17]. The accuracy of these annotations is crucial for the model's performance. Incorrect or inconsistent annotations can lead to poor training and inaccurate counting results. The model learns what constitutes an animal based on the annotations provided during training; therefore, inaccurate annotations will confuse the model and affect its ability to generalize to new images.

There are various strategies for data collection and annotation. It is recommended to collect flight data at different times of the day to capture variations in lighting, as well as in various weather conditions to ensure the robustness of the model. Specialized annotation tools, such as LabelMe , can be used to facilitate the process of labeling images and videos . The use of data augmentation techniques can also be considered to artificially increase the size of the training dataset by applying transformations to existing images (e.g., rotations, scaling changes, brightness adjustments). Finally, the possibility of collaborating with other livestock farmers or institutions to share annotated data and accelerate the model training process could be explored.

Obtaining high-quality training data for livestock counting with AI can present several challenges. The time and cost associated with collecting and annotating large datasets can be significant. In addition, specialized expertise is required to perform accurate annotations, especially in cases of occlusion or poor visibility. It is also important to consider the privacy implications when collecting data that may contain information about the location of livestock farms.

10. Conclusions and Recommendations: Integrating Technology for the Future of Livestock Counting.

In summary, livestock counting using drone images and videos with artificial intelligence represents a promising alternative to traditional methods, offering efficiency, accuracy, and the possibility of more continuous monitoring. Traditional methods, although historically relevant, present significant limitations in terms of labor intensity, susceptibility to errors, difficulty in large areas, and inconvenience for livestock . Artificial intelligence, especially through deep learning and CNNs, provides the necessary tools for automated analysis of images and videos, allowing for the accurate detection and counting of objects, including animals [10].

To implement this technology effectively, it is crucial to select the appropriate drone, considering its autonomy, camera quality (RGB and/or thermal), flight stability, and payload capacity . The choice of software and AI processing platform is also fundamental, balancing accuracy, ease of use, data handling capacity, and cost . It is imperative to ensure compliance with ANAC regulations for drone operation in the Buenos Aires region, including drone registration and obtaining the necessary permits for commercial operations . Finally, the availability of a high-quality training dataset, with a sufficient number of meticulously annotated images and videos, is essential to achieve the desired accuracy in livestock counting [51].

For those interested in adopting this technology, it is recommended to start with a pilot project on a limited scale to evaluate its effectiveness and make the necessary adjustments. The selection of hardware and software should be based on the specific needs of each livestock farm, considering pasture size, environmental conditions, and available budget. It is crucial to invest in data collection and annotation to ensure the accuracy of AI models and to consider training personnel in drone operation and software use.

The integration of artificial intelligence and drones has the potential to transform livestock management in the future, offering more efficient, accurate, and continuous monitoring. This can lead to better decision-making, increased productivity, improved animal welfare, and, ultimately, more sustainable and profitable management of livestock farms in the Buenos Aires region and beyond.

Obras citadas

1. Seeing in the Dark: Thermal Imaging Cameras are Transforming Nocturnal Wildlife Surveys, fecha de acceso: marzo 26, 2025, https://naturespy.org/blogs/journal/seeing-in-the-dark-thermal-imaging-cameras-are-transforming-nocturnal-wildlife-surveys

2. Grouping and Grazing - National Council of Teachers of Mathematics, fecha de acceso: marzo 26, 2025, https://www.nctm.org/Classroom-Resources/Illuminations/Interactives/Grouping-and-Grazing/

3. The Importance of Weighing Livestock - Adam Equipment, fecha de acceso: marzo 26, 2025, https://adamequipment.com/content/post/the-importance-of-weighing-livestock/

4. Counting Animals - African Wildlife Foundation, fecha de acceso: marzo 26, 2025, https://www.awf.org/documents/AWF_1_counting_animals.pdf

5. The Growing Need to Digitize Livestock Record Keeping Process - Folio3 AgTech, fecha de acceso: marzo 26, 2025, https://agtech.folio3.com/blogs/livestock-record-keeping/

6. Survey of Livestock Counting and Tracking Methods - NST Proceeding, fecha de acceso: marzo 26, 2025, https://nstproceeding.com/index.php/nuscientech/article/download/255/252/771

7. Livestock counting solution ready for next phase of development - evokeAG., fecha de acceso: marzo 26, 2025, https://www.evokeag.com/livestock-counting-solution-ready-for-development/

8. Artificial intelligence for livestock: a narrative review of the applications of computer vision systems and large language models for animal farming - PMC, fecha de acceso: marzo 26, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC11700613/

9. Drone Livestock Monitoring | Benefits, Tech & How it Works - Saiwa, fecha de acceso: marzo 26, 2025, https://saiwa.ai/sairone/blog/drone-livestock-monitoring/

10. Object Counting AI: Detect, Track, and Count Objects - Intelgic, fecha de acceso: marzo 26, 2025, https://intelgic.com/Object-Counting-AI-Detect-Track-and-Count-Multiple-Objects

11. Count Objects Online - Fast & Accurate! Try Free Now! - Saiwa, fecha de acceso: marzo 26, 2025, https://saiwa.ai/app/detection/count-objects/

12. Ranching Redefined: Drones for Livestock Management | Advexure, fecha de acceso: marzo 26, 2025, https://advexure.com/blogs/news/ranching-redefined-drones-for-livestock-management

13. Mastering Object Counting in Videos | by Lihi Gur Arie, PhD | TDS Archive | Medium, fecha de acceso: marzo 26, 2025, https://medium.com/data-science/mastering-object-counting-in-videos-3d49a9230bd2

14. Count Objects in Image Using Computer Vision Techniques - Labellerr, fecha de acceso: marzo 26, 2025, https://www.labellerr.com/blog/object-counting-and-sorting-using-computer-vision/

15. Whale counting in satellite and aerial images with deep learning - PMC, fecha de acceso: marzo 26, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC6776647/

16. Object Counting - viso.ai, fecha de acceso: marzo 26, 2025, https://viso.ai/application/object-counting/

17. Cattle Detection and Counting in UAV Images - Dataset Ninja, fecha de acceso: marzo 26, 2025, https://datasetninja.com/cattle-detection-and-counting-in-uav-images

18. Comparison of Object Detection Algorithms for Livestock Monitoring of Sheep in UAV images - Fraunhofer-Publica, fecha de acceso: marzo 26, 2025, https://publica.fraunhofer.de/bitstreams/54e91b8d-569d-4902-97a4-26818911fcf6/download

19. Animal Detection and Counting from UAV Images Using Convolutional Neural Networks, fecha de acceso: marzo 26, 2025, https://www.researchgate.net/publication/369083873_Animal_Detection_and_Counting_from_UAV_Images_Using_Convolutional_Neural_Networks

20. Counting animals in aerial images with a crowd counting model - TechRxiv, fecha de acceso: marzo 26, 2025, https://www.techrxiv.org/users/508208/articles/586210/master/file/data/aerial_animal_photography/aerial_animal_photography.pdf

21. Mask YOLOv7-Based Drone Vision System for Automated Cattle Detection and Counting | Artificial Intelligence and Applications - BonViewPress, fecha de acceso: marzo 26, 2025, https://ojs.bonviewpress.com/index.php/AIA/article/view/1603

22. Comparison of Different Methods of Animal Detection and Recognition on Thermal Camera Images - MDPI, fecha de acceso: marzo 26, 2025, https://www.mdpi.com/2079-9292/12/2/270

23. Crop Inspection Drones for Precision Farming - DSLRPros, fecha de acceso: marzo 26, 2025, https://www.dslrpros.com/pages/uses/crop-inspection-drones

24. Hover-view: How to Count Sheep Without Falling Asleep - Using Drones and AI to Monitor Flocks - ICAERUS, fecha de acceso: marzo 26, 2025, https://icaerus.eu/hover-view-how-to-count-sheep-without-falling-asleep-using-drones-and-ai-to-monitor-flocks/

25. Object counting Real time from drone video - Discussion - Ultralytics, fecha de acceso: marzo 26, 2025, https://community.ultralytics.com/t/object-counting-real-time-from-drone-video/613

26. The Advantages of Infrared Thermal Imaging in Wildlife Conservation and Research, fecha de acceso: marzo 26, 2025, https://www.gst-ir.net/news-events/infrared-knowledge/641.html

27. How to create a cattle counting project using yolo and drone footage from scratch - Medium, fecha de acceso: marzo 26, 2025, https://medium.com/@fernando.dijkinga/how-to-create-a-cattle-counting-project-using-yolo-and-drone-footage-from-scratch-b01f9bd0b6cd

28. Animal Counting and Detection Solution - Folio3 AI, fecha de acceso: marzo 26, 2025, https://www.folio3.ai/animal-detection/

29. Animal Counter - AGRI-FOOD.AI, fecha de acceso: marzo 26, 2025, https://agri-food.ai/animal-counter-page/

30. Agriculture Drones | Farming Drone for Crop Monitoring - JOUAV, fecha de acceso: marzo 26, 2025, https://www.jouav.com/blog/agriculture-drone.html

31. Visible vs. thermal detection: advantages and disadvantages - Lynred, fecha de acceso: marzo 26, 2025, https://www.lynred.com/blog/visible-vs-thermal-detection-advantages-and-disadvantages

32. MicaSense RedEdge-P - Multispectral and panchromatic camera - Wingtra, fecha de acceso: marzo 26, 2025, https://wingtra.com/mapping-drone-wingtraone/drone-sensors/micasense-rededge-p/

33. Enhancing Livestock Management: The Advantages of Multi-Spectral Cameras, fecha de acceso: marzo 26, 2025, https://livestockgentec.ualberta.ca/2023/10/26/enhancing-livestock-management-the-advantages-of-multi-spectral-cameras/

34. Count Livestock - CountThings.com, fecha de acceso: marzo 26, 2025, https://countthings.com/en/guides/0020

35. Crop Monitoring - Top 10 Best Drones (2025) - Fly Eye, fecha de acceso: marzo 26, 2025, https://www.flyeye.io/crop-monitoring-top10-best-drones/

36. The best camera drones in 2025: take your photography & video to the skies, fecha de acceso: marzo 26, 2025, https://www.digitalcameraworld.com/buying-guides/the-10-best-camera-drones

37. Thermal Cameras for Animal Detection - Teslong, fecha de acceso: marzo 26, 2025, https://teslong.com/blogs/news/thermal-cameras-for-animal-detection

38. Integration of thermal and RGB camera on UAV for cattle and tick-carrier identification - Texas A&M University-Corpus Christi, fecha de acceso: marzo 26, 2025, https://www.tamucc.edu/academics/honors-program/assets/documents/project-of-excellence/integration-of-thermal-and-rgb-camera-on-uav-for-cattle-and-tick-carrier-identification.pdf

39. Using Thermal Drones for Wildlife Surveys | How Do Drones Work? - Scott Country, fecha de acceso: marzo 26, 2025, https://www.scottcountry.co.uk/news/using-thermal-drones-for-wildlife-surveys

40. Thermal Game Count - Sky Senderos, fecha de acceso: marzo 26, 2025, https://www.skysenderos.com/services/thermal-game-count

41. Hot on the Trail: Can Thermal Cameras Detect Animals? - NatureSpy, fecha de acceso: marzo 26, 2025, https://naturespy.org/blogs/journal/can-thermal-cameras-detect-animals-trashed

42. 5 Advantages to Observe Wildlife at Night with Thermal Imaging - - InfiRay, fecha de acceso: marzo 26, 2025, https://www.infiray.com/5-advantages-to-observe-wildlife-at-night-with-thermal-imaging.html

43. Tracking Solution for Livestock Management | ZenaDrone Inc., fecha de acceso: marzo 26, 2025, https://www.zenadrone.com/livestock-management/

44. The Farmer's Drone Camera: RGB, Multispectral, and LIDAR Explained | Dronefly, fecha de acceso: marzo 26, 2025, https://www.dronefly.com/blogs/tech-breakdown/rgb-multispectral-and-lidar-for-farming-drones

45. Livestock Management Drones - DSLRPros, fecha de acceso: marzo 26, 2025, https://www.dslrpros.com/pages/uses/livestock-management

46. Multispectral imaging for medical and industrial machine vision systems - JAI, fecha de acceso: marzo 26, 2025, https://www.jai.com/multispectral-imaging

47. Drone Laws in Argentina | UAV Coach (2023), fecha de acceso: marzo 26, 2025, https://uavcoach.com/drone-laws-in-argentina/

48. Drone Laws in Argentina - Earth Of Drones, fecha de acceso: marzo 26, 2025, https://earthofdrones.com/latest-drone-laws-in-argentina/

49. Argentina Drone Laws (Guide 2025) | Can I Fly, fecha de acceso: marzo 26, 2025, https://canifly.in/argentina

50. Drone Laws in Argentina [Updated January 15, 2025], fecha de acceso: marzo 26, 2025, https://drone-laws.com/drone-laws-in-argentina/

51. Feasibility Analyses of Real-Time Detection of Wildlife Using UAV-Derived Thermal and RGB Images - MDPI, fecha de acceso: marzo 26, 2025, https://www.mdpi.com/2072-4292/13/11/2169

52. 6 Smart Animal Livestock Management Using Vision AI - Labellerr, fecha de acceso: marzo 26, 2025, https://www.labellerr.com/blog/smart-animal-livestock-management-use-cases-using-vision-ai/