"""A small convolutional classifier trained entirely from scratch.""" from __future__ import annotations import math from collections.abc import Sequence from torch import Tensor, nn class PlantDiseaseCNN(nn.Module): """Four convolutional stages followed by global pooling and a classifier.""" def __init__( self, num_classes: int = 39, channels: Sequence[int] = (32, 64, 128, 256), dropout: float = 0.3, ) -> None: super().__init__() if isinstance(num_classes, bool) or not isinstance(num_classes, int) or num_classes < 2: raise ValueError("num_classes must be an integer of at least 2") if len(channels) != 4 or any( isinstance(c, bool) or not isinstance(c, int) or c < 1 for c in channels ): raise ValueError("channels must contain four positive integers") if isinstance(dropout, bool) or not math.isfinite(dropout) or not 0 <= dropout < 1: raise ValueError("dropout must be finite and in [0, 1)") self.model_config = { "num_classes": num_classes, "channels": list(channels), "dropout": float(dropout), } blocks: list[nn.Module] = [] in_channels = 3 for out_channels in channels: blocks.append( nn.Sequential( nn.Conv2d(in_channels, out_channels, 3, padding=1, bias=False), nn.BatchNorm2d(out_channels), nn.ReLU(inplace=True), nn.Conv2d(out_channels, out_channels, 3, padding=1, bias=False), nn.BatchNorm2d(out_channels), nn.ReLU(inplace=True), nn.MaxPool2d(2), ) ) in_channels = out_channels self.features = nn.Sequential(*blocks) self.classifier = nn.Sequential( nn.AdaptiveAvgPool2d(1), nn.Flatten(), nn.Dropout(dropout), nn.Linear(channels[-1], num_classes), ) def forward(self, images: Tensor) -> Tensor: """Return unnormalised logits; CrossEntropyLoss applies log-softmax.""" return self.classifier(self.features(images))