调整权重存储目录。

LIU42 · Sep 29, 2024 · 34ce2bb · 34ce2bb
1 parent 1dd46d2
commit 34ce2bb
Show file tree

Hide file tree

Showing 4 changed files with 7 additions and 5 deletions.
diff --git a/.gitignore b/.gitignore
@@ -165,5 +165,5 @@ examples/outputs/*
 
 */datasets/*
 */runs/*
-*/weights/develop/*
-*/weights/deploy/*
+*/weights/*
+*/checkpoints/*
diff --git a/README.md b/README.md
@@ -51,7 +51,7 @@ pip install -r requirements.txt
 pip install onnxruntime-gpu
 ```
 
-待识别图像默认在 <u>examples/sources/</u> 下 ，识别结果默认保存在 <u>examples/outputs/</u> 下，如果不存在请先创建。将所有待识别的图像放入待识别图像目录下，要求图像尺寸为 640x480，可以在本项目 Releases 中下载我训练好的模型权重文件，解压到项目中相应的位置（位于 <u>detect/weights/deploy/</u> 和 <u>classify/weights/deploy/</u>），运行 main.py 即可。
+待识别图像默认在 <u>examples/sources/</u> 下 ，识别结果默认保存在 <u>examples/outputs/</u> 下，如果不存在请先创建。将所有待识别的图像放入待识别图像目录下，要求图像尺寸为 640x480，可以在本项目 Releases 中下载我训练好的模型权重文件，解压到项目中相应的位置（位于 <u>detect/weights/</u> 和 <u>classify/weights/</u>），运行 main.py 即可。
 
 ```bash
 python main.py
@@ -79,3 +79,5 @@ strategy: "conservative"    # 通行规则识别策略，“conservative”（
 ```bash
 pip install ultralytics
 ```
+
+
diff --git a/classify/predict.py b/classify/predict.py
@@ -12,7 +12,7 @@ def __init__(self, configs):
             providers = ['CPUExecutionProvider']
 
         self.configs = configs
-        self.session = ort.InferenceSession(f'classify/weights/deploy/classify-{self.precision}.onnx', providers=providers)
+        self.session = ort.InferenceSession(f'classify/weights/classify-{self.precision}.onnx', providers=providers)
 
     def __call__(self, image, signals):
         for signal in signals:

diff --git a/detect/predict.py b/detect/predict.py
@@ -12,7 +12,7 @@ def __init__(self, configs):
             providers = ['CPUExecutionProvider']
 
         self.configs = configs
-        self.session = ort.InferenceSession(f'detect/weights/deploy/detect-{self.precision}.onnx', providers=providers)
+        self.session = ort.InferenceSession(f'detect/weights/detect-{self.precision}.onnx', providers=providers)
 
     def __call__(self, image):
         inputs = process.preprocess(image, size=640, padding_color=127, precision=self.precision)