""" Training Script — Wrinkle Detection CNN Run this locally (GPU recommended) to train the WrinkleCNN on your dataset, then upload the weights to HuggingFace Hub. Dataset expected layout ----------------------- data/ train/ images/ *.jpg / *.png masks/ *.png (binary: 255 = wrinkle, 0 = smooth) val/ images/ masks/ Usage ----- python train_wrinkle_model.py --data_dir data --epochs 50 --batch_size 8 """ import argparse import os from pathlib import Path import cv2 import numpy as np import torch import torch.nn as nn import torch.optim as optim from torch.utils.data import Dataset, DataLoader from torchvision import transforms from PIL import Image from models.wrinkle_detector import WrinkleCNN # ── Dataset ─────────────────────────────────────────────────────────────────── class WrinkleDataset(Dataset): def __init__(self, img_dir: str, mask_dir: str, img_size: int = 256): self.img_paths = sorted(Path(img_dir).glob("*.jpg")) + \ sorted(Path(img_dir).glob("*.png")) self.mask_dir = Path(mask_dir) self.img_size = img_size self.transform = transforms.Compose([ transforms.Resize((img_size, img_size)), transforms.ToTensor(), transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]), ]) def __len__(self): return len(self.img_paths) def __getitem__(self, idx): img_path = self.img_paths[idx] mask_path = self.mask_dir / (img_path.stem + ".png") image = Image.open(img_path).convert("RGB") mask = Image.open(mask_path).convert("L").resize( (self.img_size, self.img_size), Image.NEAREST ) img_t = self.transform(image) mask_t = torch.from_numpy(np.array(mask) / 255.0).float().unsqueeze(0) return img_t, mask_t # ── Loss ────────────────────────────────────────────────────────────────────── class DiceBCELoss(nn.Module): def __init__(self, smooth: float = 1.0): super().__init__() self.smooth = smooth self.bce = nn.BCELoss() def forward(self, pred, target): bce_loss = self.bce(pred, target) pred_flat = pred.view(-1) tgt_flat = target.view(-1) inter = (pred_flat * tgt_flat).sum() dice_loss = 1.0 - (2.0 * inter + self.smooth) / \ (pred_flat.sum() + tgt_flat.sum() + self.smooth) return 0.5 * bce_loss + 0.5 * dice_loss # ── Training loop ───────────────────────────────────────────────────────────── def train(args): device = "cuda" if torch.cuda.is_available() else "cpu" print(f"Training on {device}") train_ds = WrinkleDataset( f"{args.data_dir}/train/images", f"{args.data_dir}/train/masks", args.img_size, ) val_ds = WrinkleDataset( f"{args.data_dir}/val/images", f"{args.data_dir}/val/masks", args.img_size, ) train_loader = DataLoader(train_ds, batch_size=args.batch_size, shuffle=True, num_workers=4) val_loader = DataLoader(val_ds, batch_size=args.batch_size, shuffle=False, num_workers=4) model = WrinkleCNN().to(device) criterion = DiceBCELoss() optimizer = optim.AdamW(model.parameters(), lr=args.lr, weight_decay=1e-4) scheduler = optim.lr_scheduler.CosineAnnealingLR(optimizer, T_max=args.epochs) best_val_loss = float("inf") os.makedirs(args.save_dir, exist_ok=True) for epoch in range(1, args.epochs + 1): # train model.train() train_loss = 0.0 for imgs, masks in train_loader: imgs, masks = imgs.to(device), masks.to(device) optimizer.zero_grad() preds = model(imgs) loss = criterion(preds, masks) loss.backward() optimizer.step() train_loss += loss.item() # validate model.eval() val_loss = 0.0 with torch.no_grad(): for imgs, masks in val_loader: imgs, masks = imgs.to(device), masks.to(device) preds = model(imgs) val_loss += criterion(preds, masks).item() train_loss /= len(train_loader) val_loss /= len(val_loader) scheduler.step() print(f"Epoch {epoch:3d}/{args.epochs} " f"train={train_loss:.4f} val={val_loss:.4f} " f"lr={scheduler.get_last_lr()[0]:.2e}") if val_loss < best_val_loss: best_val_loss = val_loss save_path = os.path.join(args.save_dir, "wrinkle_cnn.pth") torch.save(model.state_dict(), save_path) print(f" ✓ Saved best model → {save_path}") print(f"\nTraining complete. Best val loss: {best_val_loss:.4f}") print(f"Weights saved to: {args.save_dir}/wrinkle_cnn.pth") print("\nTo upload to HuggingFace Hub:") print(" huggingface-cli upload your-username/wrinkle-detector-weights " f"{args.save_dir}/wrinkle_cnn.pth wrinkle_cnn.pth") if __name__ == "__main__": parser = argparse.ArgumentParser() parser.add_argument("--data_dir", default="data", help="Root data directory") parser.add_argument("--save_dir", default="checkpoints",help="Where to save weights") parser.add_argument("--epochs", type=int, default=50) parser.add_argument("--batch_size",type=int, default=8) parser.add_argument("--img_size", type=int, default=256) parser.add_argument("--lr", type=float, default=1e-4) args = parser.parse_args() train(args)