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

菜品识别程序，增加向量输出到控制台的功能！

Browse Source
This commit is contained in:
zhanghuan
2025-09-11 13:39:49 +08:00
parent d9057edbd9
commit adefc3ecb2
4 changed files with 44 additions and 25 deletions
Download Patch File Download Diff File Expand all files Collapse all files
classifier/food_classifier_app.py
+37 -18
View File
@@ -66,7 +66,7 @@ class FoodClassifierApp:
# 定义图像预处理（与训练时相同） # 定义图像预处理（与训练时相同）
self.transform = transforms.Compose([ self.transform = transforms.Compose([
transforms.Resize((32, 32)), transforms.Resize((32, 32),interpolation=transforms.InterpolationMode.BILINEAR),
transforms.ToTensor(), transforms.ToTensor(),
transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225)) transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))
]) ])
@@ -114,25 +114,26 @@ class FoodClassifierApp:
image = cv2.imdecode(nparr, cv2.IMREAD_COLOR) image = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
if image is not None: if image is not None:
# image_rgb = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
return image return image
# 方法2：如果方法1失败，尝试使用PIL # # 方法2：如果方法1失败，尝试使用PIL
from PIL import Image as PILImage # from PIL import Image as PILImage
pil_image = PILImage.open(file_path) # pil_image = PILImage.open(file_path)
#
# 转换为RGB（如果是RGBA） # # 转换为RGB（如果是RGBA）
if pil_image.mode == 'RGBA': # if pil_image.mode == 'RGBA':
pil_image = pil_image.convert('RGB') # pil_image = pil_image.convert('RGB')
elif pil_image.mode == 'L': # 灰度图 # elif pil_image.mode == 'L': # 灰度图
pil_image = pil_image.convert('RGB') # pil_image = pil_image.convert('RGB')
#
# 转换为numpy数组 # # 转换为numpy数组
image_array = np.array(pil_image) # image_array = np.array(pil_image)
#
# PIL使用RGB，OpenCV使用BGR，需要转换 # # PIL使用RGB，OpenCV使用BGR，需要转换
image = cv2.cvtColor(image_array, cv2.COLOR_RGB2BGR) # image = cv2.cvtColor(image_array, cv2.COLOR_RGB2BGR)
#
return image # return image
except Exception as e: except Exception as e:
print(f"加载图片失败: {e}") print(f"加载图片失败: {e}")
@@ -549,12 +550,30 @@ class FoodClassifierApp:
pil_image = Image.fromarray(image_rgb) pil_image = Image.fromarray(image_rgb)
# 应用预处理 # 应用预处理
resize_transform = transforms.Resize((32,32),interpolation=transforms.InterpolationMode.BICUBIC)
resized_image = resize_transform(pil_image)
if isinstance(resized_image, Image.Image):
# 转换为tensor但不归一化
to_tensor = transforms.ToTensor()
resized_tensor = to_tensor(resized_image)
print(f"缩放后tensor形状: {resized_tensor.shape}")
# 打印前5个像素值（每个通道）
print("前5个像素值 (R, G, B):")
for i in range(min(5, resized_tensor.shape[1])):
r_val = resized_tensor[0, 0, i].item() * 255 # Red通道 (转换回0-255范围)
g_val = resized_tensor[1, 0, i].item() * 255 # Green通道
b_val = resized_tensor[2, 0, i].item() * 255 # Blue通道
print(f" 像素[0,{i}]: R={r_val:.2f}, G={g_val:.2f}, B={b_val:.2f}")
input_tensor = self.transform(pil_image).unsqueeze(0) # 添加batch维度 input_tensor = self.transform(pil_image).unsqueeze(0) # 添加batch维度
input_tensor = input_tensor.to(self.device) input_tensor = input_tensor.to(self.device)
# 进行预测 # 进行预测
with torch.no_grad(): with torch.no_grad():
outputs = self.model(input_tensor) outputs = self.model(input_tensor)
print('outputs',outputs)
probabilities = F.softmax(outputs, dim=1) probabilities = F.softmax(outputs, dim=1)
confidence, predicted = torch.max(probabilities, 1) confidence, predicted = torch.max(probabilities, 1)
settings/settings.py
+3 -4
View File
@@ -15,17 +15,16 @@ VAL_DATA_DIR = os.path.join(DATASET_DIR, 'val')
TEST_DATA_DIR = os.path.join(DATASET_DIR, 'test') TEST_DATA_DIR = os.path.join(DATASET_DIR, 'test')
# 模型保存路径 # 模型保存路径
MODEL_DIR = os.path.join(BASE_DIR, 'model', '05') MODEL_DIR = os.path.join(BASE_DIR, 'model', '06')
BEST_MODEL_PATH = os.path.join(MODEL_DIR, 'best_food_model.pth') BEST_MODEL_PATH = os.path.join(MODEL_DIR, 'best_food_model.pth')
TRAINING_CURVES_PATH = os.path.join(MODEL_DIR, 'training_curves.png') TRAINING_CURVES_PATH = os.path.join(MODEL_DIR, 'training_curves.png')
TRAINING_RESULTS_PATH = os.path.join(MODEL_DIR, 'training_results.txt') TRAINING_RESULTS_PATH = os.path.join(MODEL_DIR, 'training_results.txt')
# INFERENCE_BEST_MODEL_PATH = os.path.join(BASE_DIR, 'model', '03','best_food_model.pth') # INFERENCE_BEST_MODEL_PATH = os.path.join(BASE_DIR, 'model', '05','best_food_model.pth')
INFERENCE_BEST_MODEL_PATH = BEST_MODEL_PATH INFERENCE_BEST_MODEL_PATH = BEST_MODEL_PATH
# 训练参数 # 训练参数
# NUM_EPOCHS = 100 NUM_EPOCHS = 100
NUM_EPOCHS = 13
BATCH_SIZE = 32 BATCH_SIZE = 32
LEARNING_RATE = 0.001 LEARNING_RATE = 0.001
WEIGHT_DECAY = 1e-4 WEIGHT_DECAY = 1e-4
toAndroid/toAndroid.py
+2 -2
View File
@@ -7,7 +7,7 @@ from net import create_food_cnn
# 1. 初始化模型 # 1. 初始化模型
model = create_food_cnn() model = create_food_cnn()
# 2. 加载训练好的权重 # 2. 加载训练好的权重
model.load_state_dict(torch.load("../model/02/best_food_model.pth", map_location='cpu')) model.load_state_dict(torch.load("../model/06/best_food_model.pth", map_location='cpu'))
model.eval() # 设置为推理模式 model.eval() # 设置为推理模式
# 3. 创建示例输入 (假设输入是 3x224x224 的图片) # 3. 创建示例输入 (假设输入是 3x224x224 的图片)
@@ -15,4 +15,4 @@ example_input = torch.randn(1, 3, 224, 224)
# 4. 转换为 TorchScript # 4. 转换为 TorchScript
traced_script_module = torch.jit.trace(model, example_input) traced_script_module = torch.jit.trace(model, example_input)
traced_script_module.save("../model/02/best_food_model_mobile.pt") traced_script_module.save("../model/06/best_food_model_mobile.pt")
train/train_food_classifier.py
+2 -1
View File
@@ -14,8 +14,9 @@ import time
# 添加net目录到路径 # 添加net目录到路径
# sys.path.append(os.path.join(os.path.dirname(__file__), '..', 'net')) # sys.path.append(os.path.join(os.path.dirname(__file__), '..', 'net'))
# from food_net import create_food_cnn sys.path.append(os.path.join(os.path.dirname(__file__), '..'))
from net import create_food_cnn from net import create_food_cnn
# from net import create_food_cnn
from settings import settings from settings import settings
# 设置matplotlib支持中文显示 # 设置matplotlib支持中文显示