← Back to Projects Computer Vision: Real-Time Object Detection with YOLO
Train and deploy real-time object detection models with YOLO, optimized GPU inference, REST APIs, and containerized serving.
Categories
CVML
Tech Used
YOLOv8PyTorchOpenCVFine-tuningUltralyticsTensorRTONNXCUDANumPyFastAPIFlaskDocker
Problem
Computer-vision applications need object detection that remains accurate while also meeting latency, throughput, and deployment constraints.
Approach
- Trained and fine-tuned YOLOv8 detection models using data augmentation and iterative evaluation
- Optimized inference using ONNX, TensorRT, and CUDA-oriented deployment paths
- Built image and video inference workflows with OpenCV and Ultralytics
- Served predictions through API-based workflows and containerized packaging
Results
- Created a production-oriented object-detection pipeline for real-time image and video use cases
- Prepared deployment-friendly model artifacts for GPU-accelerated inference
- Built a reusable serving pattern for monitoring, inspection, analytics, and automation applications
Demo Videos