CATTLE MEDICAL DIAGNOSIS AND PREDICTION USING MACHINE LEARNING
0 likes69 views
This document describes a machine learning system to diagnose cattle health issues using images. The system uses cameras to capture images of the eyes, nose, and neck of cattle. These images are analyzed by a deep learning model trained on a dataset of healthy and infected cattle images. The model can detect common symptoms of respiratory diseases and output whether each body part appears healthy or infected. This automatic diagnosis system aims to improve cattle health monitoring compared to traditional manual methods, reducing costs for farmers and preventing disease outbreaks.
CATTLE MEDICAL DIAGNOSIS AND PREDICTION USING MACHINE LEARNING
- 1. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 09 Issue: 05 | May 2022 www.irjet.net p-ISSN: 2395-0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 3867 CATTLE MEDICAL DIAGNOSIS AND PREDICTION USING MACHINE LEARNING Harsh J. Shah1, Chirag Sharma2, Chirag Joshi3 1,2,3-Member, Young Engineer’s Club, Science Kidz Educare, Mumbai, Maharashtra, India ---------------------------------------------------------------------***--------------------------------------------------------------------- Abstract - Cattle Livestock Rearing is a major business in India with over 300 million cattle present in India. The total number of cattle throughout the world amounts to around 1 billion. The main products of this industry are milk and milk products, meats, and hides. The cattle unable to supply the former is used for transport and on farms for ploughing. The American Cattle Industry alone is worth about $77 billion. This makes the cattle sector one of the most valued industries in the world. Yet, there is not any significant system to keep health checks on these prized possessions except by workforce. In this age of reducing labour, excess workforce is expensive and does not always guarantee correct examinations. This void of a system causes annual losses of about $4-5 billion in the US alone. This can completely be avoided by using Deep Learning. With our Health Check System, we aim to surmount this challenge by capturing images of key areas in cattle and comparing them with a massive dataset of ill and healthy cattle to decide whether the animal is healthy or needs to be quarantined to prevent further spread of infection. This will provide a major boost to the cattle industry as it will narrow down the chances of a widespread infection causing losses with the help of Deep Learning. Key Words: Cattle Rearing, Deep Learning, Machine Learning, Infection Detection, Disease Prevention 1. INTRODUCTION Livestock Rearing is a massive business today with almost a billion cattle existing throughout the world. The $77 Billion Cattle Industry in America has seen a steady decline in cattle population and has suffered losses due to extreme outbursts of diseases. In 1992, there were about 120,000,000 cattle in the US. Now, it has declined to about 92 million. In 2015, about 3.9 million cattle were lost due to either non-predator or predator causes. This amounts to about $3.7 billion in losses. These deaths are dominated by non-predator causes as they account for 98% of the deaths in cattle and 89% in calves. Majorly, Respiratory Diseases make up about 23.9% of the total non-predator deaths in cattle and 26.9% in calves. A major group of respiratory diseases is the Bovine Viral Diseases. The usual symptoms include face lesions, ocular discharge, and nasal discharge. By using a highly trained prediction model, we can ease the process of infection detection and prevent any outbreaks. This will not only reduce the need for a large workforce, but also will provide a fool proof method to keep a check on livestock. 1.1 Traditional Method Steps Involved in Traditional Infection Detection: BVD is diagnosed based on samples tested at laboratories. Necroscopy findings are also considered. Body Temperature is kept under a strict vigilance and any fluctuations are noted. Milk Production is monitored. A sudden drop in production indicates infection. Nasal discharge and ocular discharge(eyes) are observed. 1.2 Problems Faced: 1. Low Infection Detection Rate. 2. Large Workforce needed. 3. Low Disease Incubation Period. 4. Risk of Outbreak. 5. Probability of Disease spreading to humans. 6. Human Intervention destabilises cattle 7. Symptoms difficult to observe. 2. SOLUTION The proposed solution is an easy-to-use software named “Cattle Infection Diagnosis,” which aims to ease the process of ensuring each cattle remains healthy. A setup of 3 cameras is installed to capture images of the designated area and send them over to the control centre aka the laptop. The three cameras are positioned at parts which are the most susceptible to display symptoms and this ensures that the best output is generated while using less resources. These body parts include the eye, the nose section, and the neck. Currently, cattle diagnosis requires dealing with hostile cattle which makes the process
- 2. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 09 Issue: 05 | May 2022 www.irjet.net p-ISSN: 2395-0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 3868 inefficient. My deep learning model aims to make the process smoother and accurate. Fig –1: Prototype Software This model can be implemented in washing sheds which will help do two tasks at one time. To diagnose a cow, the caretaker must make sure that the cow is aligned with the cameras. Once the cow is in place, he/she should trigger the cameras which will capture images and port them to the software. This will pass the image through the classification model and a result will be produced. Or else, the path of the image can be passed through and an output can be received. If the cow has ocular discharge, an output of an “Ill Eye” will be received. If the cow has a healthy nose section, a “Healthy Nose” output is received. Likewise, such outputs are received from all the sections. This helps caretakers determine whether the cattle are healthy or otherwise. Fig -2: Model (not yet implemented) 3. METHOD Fig –3: Methodology The system is built upon the “TensorFlow” machine learning library as well as the “Keras” deep neural network library in Python. The model includes three cameras and a control center (personal computer). All the classification is done through a Python code. 3. IMPLEMENTATION AND RESULTS The system aims to diagnose symptoms in the nasal, eye and the neck area Fig –4: Flowchart of the software
- 3. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 09 Issue: 05 | May 2022 www.irjet.net p-ISSN: 2395-0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 3869 In all, there are 6 possibilities: 1st Possibility: HEALTHY EYE Fig –5: Prediction 1- Healthy Eye If the cow has a healthy eye, a similar output will be displayed. Here, the path guides to the image of a healthy eye. On pressing “Enter” the image is run through 10 layers of classification and is funneled into its respective category. The image is displayed as well. A Clear Eye suggests that the cow does not have ocular discharge, a major symptom of the Bovine Respiratory Disease. 2nd Possibility: HEALTHY NASAL SECTION Fig –6: Prediction 2- Healthy Nose If the cow does not have any nasal discharge, it again eliminates a major symptom of the Bovine Disease. HEALTHY SIDE Fig –7: Prediction 3- Healthy Side If the cow has an infection-less neck region, an output of Healthy Side will be received. Any minor anomaly will also be classified as a healthy side. ILL EYE Fig –8: Prediction 4- Ill Eye If the cow has Kerato-conjunctivitis and a cloudy/foggy eye is captured, it is classified as an Ill eye. Also, if the eye is either bloody or has flesh extrusions, it will again be classified as an Ill eye. Also, a similar result will be observed if the cow has ocular discharge.
- 4. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 09 Issue: 05 | May 2022 www.irjet.net p-ISSN: 2395-0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 3870 ILL NOSE Fig –9: Prediction 5- Ill Nose If the cow shows signs of mucous discharge through nostrils, an Ill Nose output is received. ILL SIDE Fig –10: Prediction 8- Ill Side If the neck region has ringworms, lesions, or any other dermatological disease, the output received is Ill Side. CAPTURE FUNCTION: All the same outputs can be received if the “CAPTURE” and the “PROCEED” buttons are used. Fig –11: Prediction through a camera CONCLUSION: The target of this software is to provide resources to determine whether cattle are healthy or unhealthy through machine learning. It aims to help cattle owners to keep their livestock healthy without human intervention. It would be helpful to increase efficiency in infection diagnosis and prevent any contagious diseases from spreading amongst either cattle or from cattle to humans. This will be revolutionary as the number of cattle lost per year will decrease and in turn reduce losses and make the cattle Industry lucrative again. REFERENCES: https://www.simplilearn.com/keras-vs-tensorflow-vs- pytorch-article https://www.aphis.usda.gov/animal_health/nahms/gener al/downloads/cattle_calves_deathloss_2015.pdf https://www.bmc.com/blogs/deep-neural-network/ https://www.analyticsvidhya.com/blog/2022/01/image- classification-using-machine-learning/ BIOGRAPHIES: 1. HARSH J. SHAH: Technology Enthusiast & Innovator. 2. CHIRAG SHARMA: Electronics Engineer, Technical Mentor and Researcher. 3. CHIRAG JOSHI: Electronics Engineer, Technical Mentor and Researcher.