xttmp 2.3.0.7__tar.gz → 2.3.0.9__tar.gz

{xttmp-2.3.0.7/src/xttmp.egg-info → xttmp-2.3.0.9}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: xttmp
-Version: 2.3.0.7
+Version: 2.3.0.9
 Summary: eXtremely Tiny Target - Motion Perception
 Author-email: Shawn MX <mingshuoxu@hotmail.com>
 Project-URL: Homepage, https://github.com/MingshuoXu/Small-Target-Motion-Detectors

{xttmp-2.3.0.7 → xttmp-2.3.0.9}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "xttmp"
-version = "2.3.0.7"
+version = "2.3.0.9"
 description = "eXtremely Tiny Target - Motion Perception"
 readme = "README.md"
 requires-python = ">=3.8"

{xttmp-2.3.0.7 → xttmp-2.3.0.9}/src/xttmp/core/math_operator.py

@@ -114,31 +114,30 @@ class GammaDelay(BaseCore):
     Implements a gamma filter used in the lamina layer of the ESTMD neural network
     using pure PyTorch and collections.deque for efficient temporal sliding windows.
     """
-    def __init__(self, order=1, tau=1.0, kernel_len=None):
+    def __init__(self, order=1, tau=1.0):
         """
         Constructor method.
         
         Parameters:
             order (int): Order of the gamma filter (n). Default is 1.
             tau (float): Time constant of the filter (\tau).
-            kernel_len (int): Length of the filter kernel (T). 
         """
         super().__init__()
         self.order = max(1, int(order))
         self.tau = tau
-        # 如果未指定长度，默认使用 3 * tau (覆盖大部分有效权重)
-        self.kernel_len = int(3 * tau) if kernel_len is None else kernel_len
+
 
         self.setup()
 
     def setup(self):
+        _kernel_len = int(3 * self.tau)
         # 1. 预计算 Gamma 滤波器的时域权重
-        kernel = create_gamma_kernel(self.order, self.tau, self.kernel_len)
+        kernel = create_gamma_kernel(self.order, self.tau, _kernel_len)
         # 注册为 buffer，随模型自动转移设备 (如 .cuda())
         self.register_buffer('gamma_kernel', kernel)
 
         # 2. 初始化双端队列作为时序状态缓存区
-        self.buffer = deque(maxlen=self.kernel_len)
+        self.buffer = deque(maxlen=_kernel_len)
 
     def reset_buffer(self):
         """
@@ -179,15 +178,13 @@ class GammaBandPassFilter(BaseCore):
 
     def __init__(self, 
                  order1=2, tau1=3.0, 
-                 order2=6, tau2=9.0, 
-                 kernel_len=None):
+                 order2=6, tau2=9.0):
         """
         Constructor method.
         
         Parameters:
         - order1, tau1: Parameters for the excitatory (positive) Gamma filter.
         - order2, tau2: Parameters for the inhibitory (negative) Gamma filter.
-        - kernel_len: Temporal length of the filter. If None, auto-calculated.
         """
         super().__init__()
 
@@ -196,17 +193,15 @@ class GammaBandPassFilter(BaseCore):
         self.order2 = max(1, int(order2))
         self.tau2 = tau2
         
-        # 自动计算所需的历史帧缓存最大长度
-        self.kernel_len = kernel_len if kernel_len is not None else max(int(3 * tau1), int(3 * tau2))
-
         self.in_loop = False  # 默认不覆盖历史帧，直接追加
 
         self.setup()
 
     def setup(self):
+        _kernel_len = max(int(3 * self.tau1), int(3 * self.tau2))
         # 1. 预计算两个 Gamma 滤波器的权重，并补齐到相同的长度 self.T
-        k1 = create_gamma_kernel(self.order1, self.tau1, self.kernel_len)
-        k2 = create_gamma_kernel(self.order2, self.tau2, self.kernel_len)
+        k1 = create_gamma_kernel(self.order1, self.tau1, _kernel_len)
+        k2 = create_gamma_kernel(self.order2, self.tau2, _kernel_len)
 
         # 2. 算子融合 (Operator Fusion)：W_bandpass = W1 - W2
         # 直接将差值注册为模型的 buffer，前向传播只需计算一次
@@ -214,7 +209,7 @@ class GammaBandPassFilter(BaseCore):
         self.register_buffer('bandpass_kernel', bandpass_kernel)
 
         # 3. 初始化单一的高效时序状态缓存区
-        self.buffer = deque(maxlen=self.kernel_len)
+        self.buffer = deque(maxlen=_kernel_len)
 
     def reset_buffer(self):
         """

{xttmp-2.3.0.7 → xttmp-2.3.0.9}/src/xttmp/core/stmdplus_core.py

@@ -21,6 +21,8 @@ class ContrastPathway(BaseCore):
         
         self.register_buffer('T1_kernel', torch.empty(0))
 
+        self.setup()
+
     def setup(self):
         """Initialization method."""
         # 假设 create_T1_kernels 返回的是 4 个 kernel 的 List 或 NumPy array
@@ -34,9 +36,9 @@ class ContrastPathway(BaseCore):
         """
         
         # out 的形状是 (1, 4, H, W)
-        self.Opt = F.conv2d(x, self.T1_kernel, padding='same')
+        self.output = F.conv2d(x, self.T1_kernel, padding='same')
 
-        return self.Opt
+        return self.output
     
 
 class MushroomBody(BaseCore):
@@ -59,6 +61,8 @@ class MushroomBody(BaseCore):
         self.trackLens = None   # 张量: [N] (记录当前轨迹有效长度)
         self.trackPtr = None    # 张量: [N] (记录环形缓冲区的写入指针)
 
+        self.setup()
+
     def setup(self):
         self.torch_nms = AreaNMS(self.nms_size)
 
@@ -69,18 +73,18 @@ class MushroomBody(BaseCore):
         nmsLobulaOpt = self.torch_nms(maxLobulaOpt)
 
         mask_not_nms = (nmsLobulaOpt == 0)
-        mushroomBodyOpt = lobulaOpt * mask_not_nms
+        self.output = lobulaOpt * mask_not_nms
 
         maxNumber = torch.max(nmsLobulaOpt)
         if maxNumber <= 0:
             self.trackID = None
-            return mushroomBodyOpt
+            return self.output
 
         # --- 获取新检测点 ---
         newID = torch.nonzero(nmsLobulaOpt > 0).float() 
         if len(newID) == 0:
             self.trackID = None
-            return mushroomBodyOpt
+            return self.output
 
         curr_y, curr_x = newID[:, -2].long(), newID[:, -1].long()
         # all_new_contrasts 形状 [C, M], M是新目标数
@@ -180,8 +184,7 @@ class MushroomBody(BaseCore):
                     e_y = erase_coords[:, -2].long()
                     e_x = erase_coords[:, -1].long()
                     # 终极一键批量擦除
-                    mushroomBodyOpt[..., e_y, e_x] = 0
+                    self.output[..., e_y, e_x] = 0
 
-        self.Opt = mushroomBodyOpt
-        return mushroomBodyOpt
+        return self.output
 

{xttmp-2.3.0.7 → xttmp-2.3.0.9}/src/xttmp/core/vstmd_core.py

@@ -73,6 +73,11 @@ class Medulla(BaseCore):
         self.on_pathway.setup()
         self.off_pathway.setup()
 
+    def reset_buffer(self):
+        # Reset buffers for both pathways
+        self.on_pathway.reset_buffer()
+        self.off_pathway.reset_buffer()
+
     def forward(self, lamina_ON, lamina_OFF):
         """
         Process the input through the Medulla layer.

{xttmp-2.3.0.7 → xttmp-2.3.0.9}/src/xttmp/model/feedback_model.py

@@ -41,8 +41,8 @@ class FeedbackSTMD(ESTMDBackbone):
         # lamina
         'n1'        : 'lamina.order1', # Eq. (3)
         'tau1'      : 'lamina.tau1',
-        'n2'        : 'lamina.order1',
-        'tau2'      : 'lamina.tau1',
+        'n2'        : 'lamina.order2',
+        'tau2'      : 'lamina.tau2',
         # medulla
         'n3'        : ('medulla.tm1.order', 'medulla.mi1.order'), # Eq. (7)
         'tau3'      : ('medulla.tm1.tau', 'medulla.mi1.tau'), 
@@ -70,15 +70,11 @@ class FeedbackSTMD(ESTMDBackbone):
         # Customize Lobula component
         self.lobula = feedbackstmd_core.Lobula()
         
-        # Customize Lamina's GammaBankPassFilter properties
-        self.lamina.order1 = 4
-        self.lamina.tau1 = 8
-        self.lamina.order2 = 16
-        self.lamina.tau2 = 32
-
-        # Customize Medulla's Tm1 component properties
-        self.medulla.tm1.order = 9
-        self.medulla.tm1.tau = 45
+        self.set_para(n1=4, tau1=8,
+                      n2=16, tau2=32,
+                      n3=9, tau3=45)
+        
+        self.setup()
 
     def forward(self, x):
         """ MODEL_STRUCTURE Method

{xttmp-2.3.0.7 → xttmp-2.3.0.9}/src/xttmp/util/iostream.py

@@ -282,6 +282,8 @@ class FrameVisualizer:
             cv2.putText(frame, str(show_str),
                         (10, 30), cv2.FONT_HERSHEY_SIMPLEX, 0.8,
                         (0, 255, 0), 2, cv2.LINE_AA)
+
+        self.showed_frame = frame  # 保存当前帧，便于后续处理或保存
             
         # --- 视频保存 (安全检查) ---
         if self.save_output and self.video_writer is not None:

{xttmp-2.3.0.7 → xttmp-2.3.0.9/src/xttmp.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: xttmp
-Version: 2.3.0.7
+Version: 2.3.0.9
 Summary: eXtremely Tiny Target - Motion Perception
 Author-email: Shawn MX <mingshuoxu@hotmail.com>
 Project-URL: Homepage, https://github.com/MingshuoXu/Small-Target-Motion-Detectors