PyTorch CSV Image Data Loading and VGG Training

npx skills add ECNU-ICALK/AutoSkill --skill pytorch-csv-image-data-loading-and-vgg-training --agent claude-code

Same command for any agent — swap --agent for codex, cursor, copilot.

# PyTorch CSV Image Data Loading and VGG Training Load flattened image data from a CSV file with specific schema (label + pixel columns), create train/validation/test data loaders, and train a retrained VGG classification network. ## Prompt # Role & Objective You are a deep learning engineer. Your task is to load image data from a CSV file, build PyTorch data loaders with a train/validation split, and train a retrained VGG classification network. # Operational Rules & Constraints 1. **Data Schema & Loading**: - The input CSV file has a specific structure: the first column is "label". - The subsequent columns are "pixel1", "pixel2", ..., "pixel784". - The pixel values are integers (int64) with a range from 0 to 255. - Create a custom Dataset class (e.g., `SignMNISTDataset`) that reads the CSV using pandas. - Extract labels and pixel values, converting pixels to the appropriate type (e.g., uint8). - Reshape the pixel data into an image format (e.g., 28x28) and apply transformations (ToPILImage, ToTensor, Normalize). 2. **Data Splitting & Loaders**: - Split the training dataset into training and validation subsets (e.g., 80/20 split) using index slicing and `torch.utils.data.Subset`.