release: 0.3.2

README.md

@@ -201,7 +201,7 @@ from stream_infer.dispatcher import Dispatcher
 import requests
 ...
 class RequestDispatcher(Dispatcher):
-    def __init__(self, max_size: int = 120):
+    def __init__(self, max_size):
         super().__init__(max_size)
         self.sess = requests.Session()
         ...
@@ -217,14 +217,18 @@ class RequestDispatcher(Dispatcher):
 ...
 
 # Offline inference
-dispatcher = RequestDispatcher.create(offline=True, max_size=140)
+dispatcher = RequestDispatcher.create(offline=True, max_size=30)
 
 # Real-time inference
-dispatcher = RequestDispatcher.create(max_size=150)
+dispatcher = RequestDispatcher.create(max_size=15)
 ```
 
+You may have noticed that the instantiation of dispatcher differs between offline and real-time inference. This is because **in real-time inference, playback and inference are not in the same process**, and both need to share the same dispatcher, only the offline parameter has been changed, but the internal implementation uses the DispatcherManager agent.
+
 > [!CAUTION]
-> You may have noticed that the instantiation of dispatcher differs between offline and real-time inference. This is because **in real-time inference, playback and inference are not in the same process**, and both need to share the same dispatcher, only the offline parameter has been changed, but the internal implementation uses the DispatcherManager agent.
+> For the `max_size` parameter, the default value is 30, which keeps the latest 30 frames of ndarray data in the buffer. **The larger this parameter, the more memory the program occupies!**
+>
+> It is recommended to set it to `max_size = max(frame_count * (frame_step if frame_step else 1))` based on the actual inference interval.
 
 ### Inference
 
@@ -257,7 +261,7 @@ Here, we can give HeadDetectionAlgo a name to identify the running algorithm (ne
 The parameters for loading an algorithm are the framework's core functionality, allowing you to freely implement frame retrieval logic:
 
 - frame_count: The number of frames the algorithm needs to get, which is the number of frames the run() function will receive.
-- frame_step: Take 1 frame every `frame_step`, up to `frame_count` frames. (when `frame_count` is equal to 1, this parameter determines only the startup delay)
+- frame_step: Take 1 frame every `frame_step`, up to `frame_count` frames, receive 0. (when `frame_count` is equal to 1, this parameter determines only the startup delay)
 - interval: In seconds, indicating the frequency of algorithm calls, like `AnyOtherAlgo` will only be called once a minute to save resources when not needed.
 
 ### Producer
@@ -285,9 +289,11 @@ from stream_infer import Player
 
 ...
 
-player = Player(dispatcher, producer, video_path)
+player = Player(dispatcher, producer, video_path, show_progress)
 ```
 
+The `show_progress` parameter defaults to True, in which case the tqdm is used to display the progress bar. When set to False, progress is printed through the logger.
+
 ### Run
 
 Simply run the entire script through Inference's `start()`.

README.zh.md

@@ -199,7 +199,7 @@ from stream_infer.dispatcher import Dispatcher
 import requests
 ...
 class RequestDispatcher(Dispatcher):
-    def __init__(self, max_size: int = 120):
+    def __init__(self, max_size):
         super().__init__(max_size)
         self.sess = requests.Session()
         ...
@@ -215,14 +215,18 @@ class RequestDispatcher(Dispatcher):
 ...
 
 # 离线推理
-dispatcher = RequestDispatcher.create(offline=True, max_size=140)
+dispatcher = RequestDispatcher.create(offline=True, max_size=30)
 
 # 实时推理
-dispatcher = RequestDispatcher.create(max_size=150)
+dispatcher = RequestDispatcher.create(max_size=15)
 ```
 
+您可能注意到，在离线推理和实时推理下实例化 dispatcher 的方式不同，这是因为 **实时推理下播放与推理不在一个进程中** ，而两者都需要共享同一个 dispatcher，虽然只是改变了 offline 参数，但其内部实现使用了 DispatcherManager 代理。
+
 > [!CAUTION]
-> 您可能注意到，在离线推理和实时推理下实例化 dispatcher 的方式不同，这是因为 **实时推理下播放与推理不在一个进程中** ，而两者都需要共享同一个 dispatcher，虽然只是改变了 offline 参数，但其内部实现使用了 DispatcherManager 代理。
+> 对于 `max_size`参数，默认值为 30，会将最新的 30 帧 ndarray 数据存在缓冲区中，**该参数越大，程序占用的内存就越大！**
+>
+> 建议根据实际推理间隔情况设置为 `max_size = max(frame_count * (frame_step if frame_step else 1))`
 
 ### Inference
 
@@ -255,7 +259,7 @@ inference.load_algo(AnyOtherAlgo("other"), 5, 6, 60)
 而加载算法的几个参数则是框架的核心功能，让您能自由实现取帧逻辑：
 
 - frame_count：算法需要获取的帧数量，也就是最终 run() 函数中收到的 frames 数量。
-- frame_step：每隔 `frame_step` 取 1 帧，共取 `frame_count` 帧。（当 `frame_count` 为 1 时，这个参数决定的只是启动延迟）
+- frame_step：每隔 `frame_step` 取 1 帧，共取 `frame_count` 帧，可为 0。（当 `frame_count` 为 1 时，这个参数决定的只是启动延迟）
 - interval：单位秒，表示算法调用频率，如 `AnyOtherAlgo` 就只会在一分钟才调用一次，用来在不需要调用它的时候节省资源
 
 ### Producer
@@ -283,9 +287,11 @@ from stream_infer import Player
 
 ...
 
-player = Player(dispatcher, producer, video_path)
+player = Player(dispatcher, producer, video_path, show_progress)
 ```
 
+`show_progress` 参数默认为 True，此时会使用 tqdm 显示进度条，而设置为 False 时会通过 logger 打印。
+
 ### Run
 
 通过 Inference 的 `start()` 即可简单运行整个脚本

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "stream_infer"
-version = "0.3.1"
+version = "0.3.2"
 description = "Video streaming inference framework, integrating image algorithms and models for real-time/offline video structuring"
 authors = ["ZaiGie <[email protected]>"]
 maintainers = ["ZaiGie <[email protected]>"]
@@ -42,7 +42,7 @@ opencv-python-headless = ">=4.5.5.64,<=4.8.1.78"
 av = ">= 11.0.0"
 loguru = ">=0.6.0"
 streamlit = { version = ">=1.29.0", optional = true }
-
+tqdm = ">=4.62.3,<=4.66.1"
 
 [tool.poetry.extras]
 desktop = ["opencv-python"]

src/stream_infer/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "0.3.1"
+__version__ = "0.3.2"
 
 from .inference import Inference
 from .player import Player