yixiong/FoodClassifier
3
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

增加了数据增强离线的程序（可扩充数据集），将图片压缩从32*32，调整为224*224.

Browse Source
This commit is contained in:
zhanghuan
2025-09-15 17:26:14 +08:00
parent d73331010c
commit 264d01e9b9
5 changed files with 80 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files
net/food_net.py
+13 -8
View File
@@ -17,9 +17,9 @@ class FoodCNN(nn.Module):
super(FoodCNN, self).__init__()
self.use_internal_preprocess = use_internal_preprocess
# 图片预处理变换（仅在推理时使用）
# 图片预处理变换（仅在推理时使用）（32）改为（224）
self.preprocess = transforms.Compose([
transforms.Resize((32, 32),interpolation=transforms.InterpolationMode.BILINEAR),
transforms.Resize((224, 224),interpolation=transforms.InterpolationMode.BILINEAR),
transforms.ToTensor(),
transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))
])
@@ -49,7 +49,8 @@ class FoodCNN(nn.Module):
self.dropout3 = nn.Dropout2d(0.25)
# 全连接层
self.fc1 = nn.Linear(128 * 4 * 4, 512)
# self.fc1 = nn.Linear(128 * 4 * 4, 512)
self.fc1 = nn.Linear(128 * 28 * 28, 512)
self.dropout4 = nn.Dropout(0.5)
self.fc2 = nn.Linear(512, settings.NUM_CLASSES)
@@ -76,8 +77,9 @@ class FoodCNN(nn.Module):
tensor = tensor.unsqueeze(0)
# 使用与移动端相同的插值方法缩放到32x32
tensor = F.interpolate(tensor, size=(32, 32), mode='bilinear', align_corners=False)
# tensor = F.interpolate(tensor, size=(32, 32), mode='bilinear', align_corners=False)
tensor = F.interpolate(tensor, size=(224, 224), mode='bilinear', align_corners=False)
# ImageNet标准化
mean = torch.tensor([0.485, 0.456, 0.406]).view(1, 3, 1, 1)
std = torch.tensor([0.229, 0.224, 0.225]).view(1, 3, 1, 1)
@@ -100,7 +102,8 @@ class FoodCNN(nn.Module):
# x = x.float() / 255.0
# 缩放到 32x32
x = F.interpolate(x, size=(32, 32), mode='bilinear', align_corners=False)
# x = F.interpolate(x, size=(32, 32), mode='bilinear', align_corners=False)
x = F.interpolate(x, size=(224, 224), mode='bilinear', align_corners=False)
# 缩放到 32x32 - 使用 align_corners=True 来匹配 PIL 的默认行为
# x = F.interpolate(x, size=(32, 32), mode='bilinear', align_corners=True)
@@ -140,8 +143,10 @@ class FoodCNN(nn.Module):
x = self.dropout3(x)
# 展平
x = x.view(-1, 128 * 4 * 4)
# x = x.view(-1, 128 * 4 * 4)
x = x.view(-1, 128 * 28 * 28)
# x = x.view(-1, 65536)
# 全连接层
x = F.relu(self.fc1(x))
x = self.dropout4(x)