.svg)
AI Based Classification
AI-Based Classification System for Livestock and Urine Analysis
about project
During the collaboration, the client identified a need for an AI-driven system: one for automatically recognizing cattle ear tags to keep track of livestock, another for non-invasive cattle weighing, and a third for automatically recognizing and classifying particles in urine analysis results. These systems were designed to streamline processes, improve accuracy, and reduce the need for manual intervention in these areas.
Tasks
1. Train an AI Model to Recognize Cattle Ear Tags: Develop a system that can accurately identify ear tags on cattle in various formats and under different conditions to keep track of livestock.
2. Develop a Non-Invasive Cattle Weighing System: Create a system that can measure the weight of cattle without direct contact, ensuring minimal stress to the animals.
3. Create an Automated System for Recognizing and Classifying Particles in Urine Analysis: Develop an AI-based system to automatically identify and categorize particles in urine samples which can detect Cast, Crystal, Epithelia, Epithelial nuclei, Erythrocyte, Leukocyte, and Mycete
Results
1. Cattle Ear Tag Recognition: The system achieved a 96% accuracy rate in recognizing cattle ear tags, depending on the quality of the images. The non-invasive weighing system achieved 95% accuracy.
2. Urine Particle Identification: The AI system for urine analysis demonstrated over 95% accuracy in detecting and classifying types of cells based on their observed parameters (shapes, clusters)
3. Client Satisfaction: All accuracy requirements set by the client were successfully met, leading to the successful deployment of the systems.
Techstack and standards:
- Technologies/Standards:
- AI networks: Yolo, ResNet, Inception, Xception, MobileNet, VGG 16
- Image processing: CLAHE, Binarization
- Domain knowledge: AI image, cell analysis, DNA- and RNA-Analysis, Microplate reader
process
Project Implementation
Initial Task Definition
The client briefly outlined the tasks, identified potential use cases for the recognition system, and agreed on accuracy requirements.
Neural Network Analysis
Conducted an analysis of existing neural networks to determine their applicability to the specified tasks and selected the most suitable models.
System Integration
Developed scripts to integrate data with the selected neural networks, enabling data training and result generation.
Model Training
Trained the neural networks using datasets sourced from the internet, focusing on achieving the required accuracy.
Results Evaluation
Analyzed the output of the neural networks to ensure they met the accuracy standards and fine-tuned the models as needed.
review
feedback from
our customer
contacts
you have a project, and want to release it? contact us
If you hae questions or need any general information, please complete this form to request the information you need, it will be an honor to help you