cv-pipeline-builder

Build image classification pipelines by defining preprocessing (resize, normalize), applying augmentation (rotation, flip, color jitter), selecting a backbone (ResNet, EfficientNet, ViT), leveraging transfer learning from ImageNet, and configuring training and evaluation loops. This Skill automates these steps within PyTorch or TensorFlow environments integrated with SpecWeave workflows.

Yes. This Skill builds end-to-end pipelines for object detection, semantic segmentation, and image classification. It handles image preprocessing, data augmentation, CNN backbone selection, transfer learning, and production deployment across classification, detection, and segmentation use cases.

Preprocessing includes resizing and normalization to standardize inputs. Augmentation covers rotation, flipping, and color jitter to increase training robustness. These are automatically configured and applied within the pipeline before training.

Yes. Transfer learning from ImageNet is built in, allowing you to leverage pre-trained backbone architectures like ResNet and EfficientNet rather than training from scratch, reducing training time and data requirements.

This Skill includes inference deployment integration within SpecWeave workflows. After training and evaluation, the pipeline automatically configures deployment steps, enabling your model to run in production environments with experiment tracking.

Available backbones include ResNet, EfficientNet, and Vision Transformer (ViT). The Skill automatically selects and configures the appropriate architecture based on your task, streamlining backbone selection for classification, detection, and segmentation.