xttmp 2.3.0.3__tar.gz → 2.3.0.5__tar.gz

{xttmp-2.3.0.3/src/xttmp.egg-info → xttmp-2.3.0.5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: xttmp
-Version: 2.3.0.3
+Version: 2.3.0.5
 Summary: eXtremely Tiny Target - Motion Perception
 Author-email: Shawn MX <mingshuoxu@hotmail.com>
 Project-URL: Homepage, https://github.com/MingshuoXu/Small-Target-Motion-Detectors
@@ -30,7 +30,7 @@ Dynamic: license-file
 
 # Small Target Motion Detectors, Version 2.3 (XTT-MP: Extremely Tiny Target - Motion Perception)
 
-[![PyPI version](https://img.shields.io/pypi/v/XTT-MP.svg)](https://pypi.org/project/XTT-MP/) [![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](LICENSE) [![Python](https://img.shields.io/badge/python-3.8%2B-blue)](https://www.python.org/) [![PyTorch](https://img.shields.io/badge/pytorch-1.10%2B-ee4c2c)](https://pytorch.org/)
+[![PyPI version](https://img.shields.io/pypi/v/xttmp.svg)](https://pypi.org/project/xttmp/) [![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](LICENSE) [![Python](https://img.shields.io/badge/python-3.8%2B-blue)](https://www.python.org/) [![PyTorch](https://img.shields.io/badge/pytorch-1.10%2B-ee4c2c)](https://pytorch.org/)
 
 
 **XTT-MP** (Extremely Tiny Target - Motion Perception) is a natural architecture-based framework specifically designed for detecting and perceiving the motion of **extremely small targets** in complex environments.
@@ -67,13 +67,14 @@ Built with modularity and extensibility in mind, XTT-MP provides a robust suite
 - After `pip install xttmp`, use the installed code and bring your own input data, or run from a repository checkout to access the bundled examples.
 
 ### Via PyPI
-#### CPU
+#### CPU / MPS
 ```bash
 pip install xttmp[torch]
 ```
 
 #### NVIDIA GPU (CUDA 12.6)
 ```bash
+pip install xttmp
 pip install torch torchvision --index-url https://download.pytorch.org/whl/cu126
 ```
 

{xttmp-2.3.0.3 → xttmp-2.3.0.5}/README.md

@@ -1,6 +1,6 @@
 # Small Target Motion Detectors, Version 2.3 (XTT-MP: Extremely Tiny Target - Motion Perception)
 
-[![PyPI version](https://img.shields.io/pypi/v/XTT-MP.svg)](https://pypi.org/project/XTT-MP/) [![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](LICENSE) [![Python](https://img.shields.io/badge/python-3.8%2B-blue)](https://www.python.org/) [![PyTorch](https://img.shields.io/badge/pytorch-1.10%2B-ee4c2c)](https://pytorch.org/)
+[![PyPI version](https://img.shields.io/pypi/v/xttmp.svg)](https://pypi.org/project/xttmp/) [![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](LICENSE) [![Python](https://img.shields.io/badge/python-3.8%2B-blue)](https://www.python.org/) [![PyTorch](https://img.shields.io/badge/pytorch-1.10%2B-ee4c2c)](https://pytorch.org/)
 
 
 **XTT-MP** (Extremely Tiny Target - Motion Perception) is a natural architecture-based framework specifically designed for detecting and perceiving the motion of **extremely small targets** in complex environments.
@@ -37,13 +37,14 @@ Built with modularity and extensibility in mind, XTT-MP provides a robust suite
 - After `pip install xttmp`, use the installed code and bring your own input data, or run from a repository checkout to access the bundled examples.
 
 ### Via PyPI
-#### CPU
+#### CPU / MPS
 ```bash
 pip install xttmp[torch]
 ```
 
 #### NVIDIA GPU (CUDA 12.6)
 ```bash
+pip install xttmp
 pip install torch torchvision --index-url https://download.pytorch.org/whl/cu126
 ```
 

{xttmp-2.3.0.3 → xttmp-2.3.0.5}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "xttmp"
-version = "2.3.0.3"
+version = "2.3.0.5"
 description = "eXtremely Tiny Target - Motion Perception"
 readme = "README.md"
 requires-python = ">=3.8"

{xttmp-2.3.0.3 → xttmp-2.3.0.5}/src/xttmp/core/math_operator.py

@@ -102,7 +102,9 @@ class GaussianBlur(BaseCore):
         weight = self.blur_kernel.expand(C, 1, self.kernel_size, self.kernel_size)
 
         # 使用深度可分离卷积（groups=C），每个通道独立进行高斯模糊
-        return F.conv2d(x, weight, padding='same', groups=C)
+        self.output = F.conv2d(x, weight, padding='same', groups=C)
+
+        return self.output
     
 
 class GammaDelay(BaseCore):
@@ -161,7 +163,9 @@ class GammaDelay(BaseCore):
         else:
             self.buffer.append(x)
 
-        return compute_temporal_conv_inplace(self.buffer, self.gamma_kernel)
+        self.output = compute_temporal_conv_inplace(self.buffer, self.gamma_kernel)
+
+        return self.output
 
 
 class GammaBandPassFilter(BaseCore):
@@ -236,7 +240,9 @@ class GammaBandPassFilter(BaseCore):
         else:
             self.buffer.append(x)
 
-        return compute_temporal_conv_inplace(self.buffer, self.bandpass_kernel) 
+        self.output = compute_temporal_conv_inplace(self.buffer, self.bandpass_kernel)
+
+        return self.output
 
 
 class SpatialInhibition(BaseCore):
@@ -304,4 +310,6 @@ class SpatialInhibition(BaseCore):
         weight = self.kernel.expand(C, 1, self.kernel_size, self.kernel_size)
 
         # groups=C 表示进行深度可分离卷积（Depthwise Convolution），每个通道独立滤波        
-        return F.relu(F.conv2d(x, weight, padding='same', groups=C))
+        self.output = F.relu(F.conv2d(x, weight, padding='same', groups=C))
+
+        return self.output

{xttmp-2.3.0.3 → xttmp-2.3.0.5}/src/xttmp/demo/inference_gui.py

@@ -18,9 +18,6 @@ from xttmp.util.iostream import ( # type: ignore
                 FrameIterator,
                 FrameVisualizer,
             )
-from xttmp.api import instancing_model
-
-
 # configure logging
 logging.basicConfig(level=logging.INFO,
                     format='%(asctime)s - %(levelname)s - %(message)s')
@@ -34,7 +31,6 @@ class StmdGui:
         self.FrameIterator = FrameIterator
         self.FrameVisualizer = FrameVisualizer
         self.post_processor = None
-        self.instancing_model = instancing_model
 
     def _get_user_input(self) -> tuple:
         """ get user input """
@@ -89,7 +85,9 @@ class StmdGui:
             self.post_processor = post_processor
             self.device = device
             reader = self._create_frame_reader(opt1, opt2)
-            model = self.instancing_model(model_name, device)
+            
+            from xttmp.api import instancing_model
+            model = instancing_model(model_name, device)
 
             visualizer = self.FrameVisualizer(
                 window_name=model_name,
@@ -114,7 +112,7 @@ class StmdGui:
                 run_time = time.perf_counter() - time_start
 
                 post_res = post_processor(result['response'], result.get('direction'))
-                show_str = f'{self.device} : {run_time*1000:.1f} ms'
+                show_str = f'{self.device.upper()} : {run_time*1000:.1f} ms'
                 if not visualizer.update(color_img, result=post_res, show_str=show_str):
                     break
 

{xttmp-2.3.0.3 → xttmp-2.3.0.5}/src/xttmp/main.py

@@ -2,14 +2,6 @@ from pathlib import Path
 import subprocess
 import sys
 
-try:
-    import torch
-except ImportError:
-    raise ImportError(
-        "Please install PyTorch first. "
-        "See https://pytorch.org/get-started/locally/"
-    )
-
 def main():
 	script_path = Path(__file__).resolve().parent / 'demo' / 'inference_gui.py'
 	result = subprocess.run([sys.executable, str(script_path)], check=False)

{xttmp-2.3.0.3 → xttmp-2.3.0.5}/src/xttmp/model/haarstmd.py

@@ -4,6 +4,7 @@ from ..core import haarstmd_core
 from .backbone import ESTMDBackbone
 from ..util.compute_module import compute_response, compute_direction
 
+
 class HaarSTMD(ESTMDBackbone):
     ''' HaarSTMD: Advancing small target motion detection in dim light
     

{xttmp-2.3.0.3 → xttmp-2.3.0.5}/src/xttmp/util/iostream.py

@@ -451,7 +451,7 @@ class ModelSelectorGUI:
         self.root = root
 
     def create_gui(self, modelList):
-        self.modelLabel = ttk.Label(self.root, text="Select a model:", width = 15)
+        self.modelLabel = ttk.Label(self.root, text="Select A Model:", width = 15)
         self.modelLabel.grid(row=0, column=0, padx=10, pady=10)
 
         self.modelCombobox = ttk.Combobox(self.root, values=modelList, width = 25)
@@ -596,14 +596,14 @@ class PostProcessingSelectorGUI:
         self.selectedOption = tk.IntVar(value=2)
 
         self.dotLabel = ttk.Radiobutton(self.root, 
-                                        text='dot output', 
+                                        text='STMD Output', 
                                         variable=self.selectedOption,
                                         value=2, 
                                         command=self.select_dot)
         self.dotLabel.grid(row=5, column=1, padx=10, pady=10, sticky="w")
         
         self.bboxLabel = ttk.Radiobutton(self.root, 
-                                        text='bbox output', 
+                                        text='Bbox Output', 
                                         variable=self.selectedOption,
                                         value=1, 
                                         command=self.select_bbox)
@@ -644,14 +644,20 @@ class PostProcessingSelectorGUI:
             raise ValueError(f"Unknown output type: {self.output_type}")
 
     def update_show_threshold(self, *args):
-        value = float(self.showThresholdVar.get())
+        try:
+            value = float(self.showThresholdVar.get())
+        except ValueError:
+            value = 0.0
         self.show_threshold = min(max(value, 0.0), 1.0)  # 确保在 [0.0, 1.0] 范围内
         
     def update_top_num(self, *args):
-        value = int(self.getTopNumVar.get())
-        self.top_num = max(value, 1)  # 确保 top_num 至少为 1
-
+        try:
+            value = int(self.getTopNumVar.get())
+        except ValueError:
+            # 如果输入为空或者包含非数字字符，赋予一个默认值
+            value = 0
 
+        self.top_num = max(value, 1)  # 确保 top_num 至少为 1
 
 
 class DeviceSelectorGUI:
@@ -667,26 +673,37 @@ class DeviceSelectorGUI:
         if torch.cuda.is_available():
             self.selectedOption.set(2)
             self.device = "cuda"
-
-        self.cpuLabel = ttk.Radiobutton(self.root, 
-                                        text='CPU', 
-                                        variable=self.selectedOption,
-                                        value=1, 
-                                        command=self.select_cpu)
-        self.cpuLabel.grid(row=8, column=1, padx=10, pady=10, sticky="w")
         
-        self.gpuLabel = ttk.Radiobutton(self.root, 
-                                        text='GPU', 
+        self.device_frame = ttk.Frame(self.root)
+        self.device_frame.grid(row=8, column=1, columnspan=2, padx=10, pady=10, sticky="w")
+
+        self.cpuLabel = ttk.Radiobutton(self.device_frame, 
+                                text='CPU', 
+                                variable=self.selectedOption,
+                                value=1, 
+                                command=self.select_cpu)
+        # 使用 pack(side="left") 可以让它们在 frame 内从左到右水平挨着排列
+        self.cpuLabel.pack(side="left", padx=(0, 20)) # 右侧留 20px 间距
+
+        self.cudaLabel = ttk.Radiobutton(self.device_frame, 
+                                        text='CUDA', 
                                         variable=self.selectedOption,
                                         value=2, 
-                                        command=self.select_gpu)
-        self.gpuLabel.grid(row=8, column=2, padx=10, pady=10, sticky="w")
+                                        command=self.select_cuda)
+        self.cudaLabel.pack(side="left", padx=(0, 20))
+
+        self.mpsLabel = ttk.Radiobutton(self.device_frame, 
+                                        text='MPS', 
+                                        variable=self.selectedOption,
+                                        value=3, 
+                                        command=self.select_mps)
+        self.mpsLabel.pack(side="left")
 
     def select_cpu(self):
         self.selectedOption.set(1)
         self.device = "cpu"
 
-    def select_gpu(self):
+    def select_cuda(self):
         if torch.cuda.is_available():
             self.selectedOption.set(2)
             self.device = "cuda"
@@ -694,6 +711,14 @@ class DeviceSelectorGUI:
             messagebox.showinfo("Message title", "CUDA is not available. Please select CPU.")
             self.select_cpu()
 
+    def select_mps(self):
+        if torch.backends.mps.is_available():
+            self.selectedOption.set(3)
+            self.device = "mps"
+        else:
+            messagebox.showinfo("Message title", "Metal Performance Shaders (MPS) is not available. Please select CPU.")
+            self.select_cpu()
+
 
 class XTTMP_GUI:
     def __init__(self, root):

{xttmp-2.3.0.3 → xttmp-2.3.0.5/src/xttmp.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: xttmp
-Version: 2.3.0.3
+Version: 2.3.0.5
 Summary: eXtremely Tiny Target - Motion Perception
 Author-email: Shawn MX <mingshuoxu@hotmail.com>
 Project-URL: Homepage, https://github.com/MingshuoXu/Small-Target-Motion-Detectors
@@ -30,7 +30,7 @@ Dynamic: license-file
 
 # Small Target Motion Detectors, Version 2.3 (XTT-MP: Extremely Tiny Target - Motion Perception)
 
-[![PyPI version](https://img.shields.io/pypi/v/XTT-MP.svg)](https://pypi.org/project/XTT-MP/) [![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](LICENSE) [![Python](https://img.shields.io/badge/python-3.8%2B-blue)](https://www.python.org/) [![PyTorch](https://img.shields.io/badge/pytorch-1.10%2B-ee4c2c)](https://pytorch.org/)
+[![PyPI version](https://img.shields.io/pypi/v/xttmp.svg)](https://pypi.org/project/xttmp/) [![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](LICENSE) [![Python](https://img.shields.io/badge/python-3.8%2B-blue)](https://www.python.org/) [![PyTorch](https://img.shields.io/badge/pytorch-1.10%2B-ee4c2c)](https://pytorch.org/)
 
 
 **XTT-MP** (Extremely Tiny Target - Motion Perception) is a natural architecture-based framework specifically designed for detecting and perceiving the motion of **extremely small targets** in complex environments.
@@ -67,13 +67,14 @@ Built with modularity and extensibility in mind, XTT-MP provides a robust suite
 - After `pip install xttmp`, use the installed code and bring your own input data, or run from a repository checkout to access the bundled examples.
 
 ### Via PyPI
-#### CPU
+#### CPU / MPS
 ```bash
 pip install xttmp[torch]
 ```
 
 #### NVIDIA GPU (CUDA 12.6)
 ```bash
+pip install xttmp
 pip install torch torchvision --index-url https://download.pytorch.org/whl/cu126
 ```