About Workshop
Aim
What Participants Will Learn
-
Deconstruct Architectures: Analyze the structural evolution from standard CNNs to YOLOv10's consistent dual assignments and YOLO11's C3k2 blocks.
-
Execute Zero-Shot Inference: Instantiate pre-trained models and establish baseline performance metrics on dense, complex proxy datasets.
-
Optimize Custom Training Loops: Format domain-specific data (e.g., industrial defects) and initiate parallel, distributed fine-tuning while monitoring live hardware telemetry.
-
Profile Hardware Stress: Track VRAM consumption, GPU utilization spikes, and loss degradation to prevent compute overruns.
-
Simulate Production Constraints: Export fine-tuned weights to optimized inference engines (.onnx and .engine) to test absolute edge device latency.
Structure
- Evolution of object detection: From fundamental CNNs to the YOLO family
- The end-to-end vs. versatility architectural debate
- YOLOv10 mechanics: Consistent dual assignments and NMS bottleneck elimination
- YOLO11 mechanics: C3k2 blocks and C2PSA spatial attention modules
- Setting up the environment: CUDA, Python, and PyTorch dependencies
- Configuring the Ultralytics Python API pipeline
- Instantiating pre-trained yolov10n.pt and yolo11n.pt architectures
- Running zero-shot inference on standard proxy datasets
- Understanding the YOLO YAML dataset standard and annotation formats
- Data augmentation strategies for robust feature extraction
- Optimizing gradient flow for faster convergence in YOLO11
- Efficiency versus accuracy trade-offs in YOLOv10's block design
- Formatting specialized datasets (e.g., MVTec AD for industrial defects)
- Initiating parallel distributed training loops for YOLOv10-S and YOLO11-S
- Tracking live metrics (Loss, Precision, Recall, mAP@50)
- Why standard mAP alone is insufficient for production machine learning
- Understanding computational graph optimizations in inference engines
- Architectural differences between .pt, .onnx, and .engine formats
- Exporting fine-tuned PyTorch weights to ONNX
- Compiling TensorRT engines to simulate real-world edge device constraints
- Processing a continuous 1,000-frame video stream for robust inference testing
- Capturing strict production telemetry (FPS, Peak Memory, Latency Variance)
- Mapping exact speed-accuracy tradeoffs using comparative visualizations
- 🧰 Tools Covered: ONNX, TensorRT, Matplotlib, Seaborn, Google Colab
Important Dates
Registration Ends
Workshop Dates
What You Will Gain
Outcomes
-
Architectural Mastery: The ability to empirically justify the selection of YOLOv10 vs. YOLO11 based on specific speed, accuracy, and parameter constraints.
-
Publication-Ready Assets: The capacity to generate rigorous data visualizations mapping exact speed-accuracy tradeoffs (mAP@50-95 vs. FLOPs) for academic journals or thesis defenses.
-
End-to-End MLOps Pipelines: A complete, functional codebase that takes raw datasets through PyTorch training and straight into TensorRT deployment.
-
Enterprise-Grade Profiling: The practical skills to capture strict production telemetry (FPS, Peak Memory, Latency Variance) over continuous video streams, ensuring safe and cost-effective real-world deployments.
Who Should Attend
-
PhD Scholars & Researchers: Looking to integrate state-of-the-art vision architectures into their thesis or generate publication-ready comparative metrics.
-
Industry ML Professionals & Computer Vision Engineers: Tasked with optimizing model performance, reducing latency, and deploying real-time vision pipelines to edge devices.
-
Faculty Members & Academicians: Seeking to update university curricula with the latest industry-standard practices in MLOps and object detection.