Computer Vision Model Training Diagnostics & Optimization
Debug and optimize computer vision model training with systematic loss curve and convergence analysi
What You Can Do
You can analyze computer vision training pipelines to identify root causes of poor convergence, overfitting, or dataset issues. Claude examines loss curves, gradient flow, dataset balance, and hyperparameter sensitivity to generate actionable optimization recommendations. This transforms opaque training runs into diagnosed problems with clear fixes.
Features
Example Output
Loss Curve Analysis
- 📊 Training Loss Pattern: Oscillating divergence detected
- Epoch 0-10: Smooth descent (good)
- Epoch 10-25: Erratic spikes, overall upward trend
- Diagnosis: Learning rate too high OR batch size too small
- Recommendation: Reduce learning rate by 50% or increase batch size to 128
Dataset Quality Report
- 🔍 Class Distribution
- Class A: 5,200 samples (65%)
- Class B: 1,800 samples (22%)
- Class C: 600 samples (13%)
- Issue: Severe class imbalance detected
- Fix: Apply weighted loss (weight=[1.0, 2.9, 8.7]) or use SMOTE
Hyperparameter Tuning Plan
- ⚙️ Recommended Adjustments
1. Learning Rate: 0.001 → 0.0005 (reduce due to oscillation)
2. Batch Size: 32 → 64 (stabilize gradients)
3. Add Weight Decay: 1e-4 (reduce overfitting)
4. Schedule: Use CosineAnnealingLR over 100 epochs
What's Included
- SKILL.md: Core diagnostic skill with training analysis workflows
- Loss Curve Diagnostic Template: Structured format for loss pattern analysis
- Dataset Quality Checklist: Class balance, label errors, augmentation validation
- Hyperparameter Tuning Workflow: Sensitivity analysis and recommendation framework
- Training Report Template: Executive summary with diagnosis and fixes
- Gradient Flow Analysis Guide: Debugging exploding/vanishing gradients
- Checkpoint Selection Worksheet: Choose optimal model from saved checkpoints
Who It's For
- Machine learning engineers building production CV models
- Computer vision researchers debugging training experiments
- Data scientists optimizing model convergence and accuracy
- Deep learning practitioners troubleshooting training failures
- ML Ops engineers diagnosing training pipeline bottlenecks
Best For
- Diagnosing loss divergence, oscillation, and training failures
- Identifying and fixing class imbalance and label quality issues
- Optimizing hyperparameters (learning rate, batch size, regularization)
- Detecting overfitting and choosing early stopping points
- Analyzing gradient flow and activation distributions for pathological behavior
You might also like
Rapidly analyze production incidents by correlating observability signals—logs, metrics, traces, and events—to identify the true root cause instead of symptoms. You'll reconstruct incident timelines, quantify impact scope, and generate both immediate remediation steps and long-term prevention measures.
Claude analyzes your code diffs and generates standardized commit messages that follow the Conventional Commits specification. The skill automatically determines the correct commit type, scope, and description based on the changes you've made, ensuring your messages are parseable by automation tools while remaining human-readable for code reviewers.
Transform schema and data changes into production-ready migration packages that protect data integrity and minimize downtime. You can handle column additions, renames, type conversions, and backfills using proven patterns like expand-contract and shadow tables. The skill generates explicit rollback scripts, validation checklists, and phased execution plans that keep applications stable throughout the migration.
You can submit your Solidity smart contracts for comprehensive security analysis, vulnerability detection, and gas optimization recommendations. Claude reviews your code against known attack vectors, design patterns, and best practices—identifying reentrancy risks, overflow/underflow issues, access control flaws, and inefficient operations. You receive actionable fix recommendations with explanations of the security implications and gas savings potential.
You can design end-to-end NGS pipelines tailored to your experiment type (WGS, WES, RNA-seq, targeted), troubleshoot pipeline failures by analyzing QC metrics and logs, and optimize performance by fine-tuning parameters and identifying bottlenecks. This skill handles the full workflow from raw sequencing data quality assessment through variant interpretation and annotation.
This skill guides you through designing Kalman filters for your specific state estimation and sensor fusion problems. You'll learn to diagnose filter divergence, tune covariance matrices (Q and R), and validate observability—turning mathematical theory into production-ready filters. The skill provides step-by-step workflows, diagnostic tools, and practical templates for common applications like GPS/IMU fusion, SLAM, and aircraft navigation.
You can rapidly identify root causes of training failures, data pipeline breakdowns, and model serving issues by analyzing logs, metrics, and infrastructure state. Claude examines your system events, resource utilization patterns, and error traces to pinpoint performance bottlenecks, data quality problems, and deployment misconfigurations—then recommends targeted remediation steps.
You can accelerate your entire antenna design workflow—from initial concept to optimized specifications—by leveraging Claude's knowledge of electromagnetic theory, antenna types, and design trade-offs. Claude helps you parameterize designs, identify optimal configurations for your frequency band and application, and prepare detailed briefs for simulation and manufacturing. You'll generate design reviews, impedance-matching solutions, and validation checklists without context-switching to multiple tools.