Cv Classification
Best practices for image classification tasks.
Install
npx skills add aiming-lab/AutoResearchClaw --skill cv-classification -g -a claude-code -y- Repository
- aiming-lab/AutoResearchClaw14.8k
- Category
- Media
- Updated
- 3mo ago
- License
- MIT
- Context cost
- Lightabout 183 tokens
- Runs scripts
- No, instructions only
What it does
Best practices for image classification tasks. Use when working on CIFAR, ImageNet, or other classification benchmarks.
Image Classification Best Practice
Architecture selection:
- Small scale (CIFAR-10/100): ResNet-18/34, WideResNet, Simple ViT
- Medium scale: ResNet-50, EfficientNet-B0/B1, DeiT-Small
- Large scale: ViT-B/16, ConvNeXt, Swin Transformer
Training recipe:
- Optimizer: AdamW (lr=1e-3 to 3e-4) or SGD (lr=0.1 with cosine decay)
- Weight decay: 0.01-0.1 for AdamW, 5e-4 for SGD
- Data augmentation: RandomCrop, RandomHorizontalFlip, Cutout/CutMix
- Warmup: 5-10 epochs linear warmup for transformers
- Batch size: 128-256 for CNNs, 512-1024 for ViTs (if memory allows)
Standard benchmarks:
- CIFAR-10: ~96% (ResNet-18), ~97% (WideResNet)
- CIFAR-100: ~80% (ResNet-18), ~84% (WideResNet)
- ImageNet: ~76% (ResNet-50), ~81% (ViT-B/16)
