下載 Azure Kinect Body 追蹤 SDKDownload Azure Kinect Body Tracking SDK

本檔提供安裝每個 Azure Kinect Body 追蹤 SDK 版本的連結。This document provides links to install each version of the Azure Kinect Body Tracking SDK.

Azure Kinect 主體追蹤 SDK 內容Azure Kinect Body Tracking SDK contents

  • 使用 Azure Kinect DK 建立主體追蹤應用程式的標頭和程式庫。Headers and libraries to build a body tracking application using the Azure Kinect DK.
  • 使用 Azure Kinect DK 的主體追蹤應用程式所需的可轉散發 Dll。Redistributable DLLs needed by body tracking applications using the Azure Kinect DK.
  • 範例主體追蹤應用程式。Sample body tracking applications.
版本Version 下載Download
1.0.11.0.1 msi nugetmsi nuget
1.0.01.0.0 msi nugetmsi nuget
0.9.50.9.5 msi nugetmsi nuget
0.9.40.9.4 msi nugetmsi nuget
0.9.30.9.3 msi nugetmsi nuget
0.9.20.9.2 msi nugetmsi nuget
0.9.10.9.1 msi nugetmsi nuget
0.9.00.9.0 msi nugetmsi nuget

Linux 安裝指示Linux installation instructions

目前,唯一支援的發行版本為 Ubuntu 18.04。Currently, the only supported distribution is Ubuntu 18.04. 若要要求支援其他發行版本,請參閱此頁面To request support for other distributions, see this page.

首先,您必須遵循這裡的指示,設定 Microsoft 的套件存放庫First, you'll need to configure Microsoft's Package Repository, following the instructions here.

libk4abt<major>.<minor>-dev 套件包含要針對 libk4abt 建置的標頭和 CMake 檔案。The libk4abt<major>.<minor>-dev package contains the headers and CMake files to build against libk4abt. libk4abt<major>.<minor>封裝包含執行相依于和範例檢視器的可執行檔所需的共用物件 libk4abtThe libk4abt<major>.<minor> package contains the shared objects needed to run executables that depend on libk4abt as well as the example viewer.

基本教學課程需要 libk4abt<major>.<minor>-dev 套件。The basic tutorials require the libk4abt<major>.<minor>-dev package. 若要安裝它,請執行To install it, run

sudo apt install libk4abt1.0-dev

如果命令成功,SDK 就可供使用。If the command succeeds, the SDK is ready for use.

注意

安裝 SDK 時,請記住您安裝的路徑。When installing the SDK, remember the path you install to. 例如,"C:\Program Files\Azure Kinect Body 追蹤 SDK 1.0.0"。For example, "C:\Program Files\Azure Kinect Body Tracking SDK 1.0.0". 您會在此路徑的文章中找到參考的範例。You will find the samples referenced in articles in this path. 本文追蹤範例位於 Azure Kinect 範例存放庫中的 [內文追蹤-範例] 資料夾中。Body tracking samples are located in the body-tracking-samples folder in the Azure-Kinect-Samples repository. 您會在這裡的文章中找到參考的範例。You will find the samples referenced in articles here.

變更記錄Change log

1.0.1 版v1.0.1

  • [Bug 修正]修正從 Windows 組建19025或更新版本上的路徑載入 onnxruntime.dll 時,SDK 損毀的問題:連結[Bug Fix] Fix issue that the SDK crashes if loading onnxruntime.dll from path on Windows build 19025 or later: Link

1.0.0 版v1.0.0

  • 特徵將 c # 包裝函式新增至 msi 安裝程式。[Feature] Add C# wrapper to the msi installer.
  • [Bug 修正]修正無法正確偵測到 head 旋轉的問題:連結[Bug Fix] Fix issue that the head rotation cannot be detected correctly: Link
  • [Bug 修正]修正 CPU 使用量在 Linux 電腦上到達100% 的問題:連結[Bug Fix] Fix issue that the CPU usage goes up to 100% on Linux machine: Link
  • 範例將兩個範例新增至範例存放庫。[Samples] Add two samples to the sample repo. 範例1示範如何將內容追蹤結果從深度空間轉換成色彩空間連結;範例2示範如何偵測 floor 平面連結Sample 1 demonstrates how to transform body tracking results from the depth space to color space Link; sample 2 demonstrates how to detect floor plane Link

v 0.9。5v0.9.5

  • 特徵C # 支援。[Feature] C# support. C # 包裝函式封裝在 nuget 套件中。C# wrapper is packed in the nuget package.
  • 特徵多重追蹤程式支援。[Feature] Multi-tracker support. 允許建立多個追蹤器。Creating multiple trackers is allowed. 現在,使用者可以建立多個追蹤器來追蹤來自不同 Azure Kinect 裝置的主體。Now user can create multiple trackers to track bodies from different Azure Kinect devices.
  • 特徵CPU 模式的多執行緒處理支援。[Feature] Multi-thread processing support for CPU mode. 以 CPU 模式執行時,會使用所有核心來最大化速度。When running on CPU mode, all cores will be used to maximize the speed.
  • 特徵加入 gpu_device_idk4abt_tracker_configuration_t 結構。[Feature] Add gpu_device_id to k4abt_tracker_configuration_t struct. 允許使用者指定預設值以外的 GPU 裝置,以執行主體追蹤演算法。Allow users to specify GPU device that is other than the default one to run the body tracking algorithm.
  • [Bug 修正/中斷變更]修正聯合名稱中的打字錯誤。[Bug Fix/Breaking Change] Fix typo in a joint name. 將 [聯合名稱] 從變更 K4ABT_JOINT_SPINE_NAVALK4ABT_JOINT_SPINE_NAVELChange joint name from K4ABT_JOINT_SPINE_NAVAL to K4ABT_JOINT_SPINE_NAVEL.

v 0.9。4v0.9.4

  • 特徵新增手接頭支援。[Feature] Add hand joints support. SDK 會針對每一手勢提供三個額外接點的資訊:手、HANDTIP、拇指。The SDK will provide information for three additional joints for each hand: HAND, HANDTIP, THUMB.
  • 特徵為每個偵測到的接點新增預測信賴等級。[Feature] Add prediction confidence level for each detected joints.
  • 特徵新增 CPU 模式支援。[Feature] Add CPU mode support. 藉由變更 cpu_only_mode 中的值 k4abt_tracker_configuration_t ,現在 SDK 可以在 CPU 模式上執行,而不需要使用者具備強大的圖形配接器。By changing the cpu_only_mode value in k4abt_tracker_configuration_t, now the SDK can run on CPU mode which doesn't require the user to have a powerful graphics card.

v 0.9。3v0.9.3

  • 特徵發行新的 DNN 模型 dnn_model_2_0. onnx,其主要會改善主體追蹤的健全性。[Feature] Publish a new DNN model dnn_model_2_0.onnx, which largely improves the robustness of the body tracking.
  • 特徵預設停用時態性平滑處理。[Feature] Disable the temporal smoothing by default. 追蹤的接點會更有回應。The tracked joints will be more responsive.
  • 特徵改善主體索引對應的精確度。[Feature] Improve the accuracy of the body index map.
  • [Bug 修正]修正感應器方向設定無法生效的 bug。[Bug Fix] Fix bug that the sensor orientation setting is not effective.
  • [Bug 修正]將 body_index_map 類型從 K4A_IMAGE_FORMAT_CUSTOM 變更為 K4A_IMAGE_FORMAT_CUSTOM8。[Bug Fix] Change the body_index_map type from K4A_IMAGE_FORMAT_CUSTOM to K4A_IMAGE_FORMAT_CUSTOM8.
  • [已知問題]兩個關閉主體可能會合並成單一實例區段。[Known Issue] Two close bodies may merge to single instance segment.

v 0.9。2v0.9.2

  • [重大變更]更新以相依于最新的 Azure Kinect 感應器 SDK 1.2.0。[Breaking Change] Update to depend on the latest Azure Kinect Sensor SDK 1.2.0.
  • [API 變更] k4abt_tracker_create函式會開始接受 k4abt_tracker_configuration_t 輸入。[API Change] k4abt_tracker_create function will start to take a k4abt_tracker_configuration_t input.
  • [API 變更]將 k4abt_frame_get_timestamp_usec API 變更為, k4abt_frame_get_device_timestamp_usec 以更明確且與感應器 SDK 1.2.0 一致。[API Change] Change k4abt_frame_get_timestamp_usec API to k4abt_frame_get_device_timestamp_usec to be more specific and consistent with the Sensor SDK 1.2.0.
  • 特徵允許使用者在建立追蹤器時指定感應器掛接方向,以在不同的角度進行裝載時,達到更精確的內文追蹤結果。[Feature] Allow users to specify the sensor mounting orientation when creating the tracker to achieve more accurate body tracking results when mounting at different angles.
  • 特徵提供新的 API k4abt_tracker_set_temporal_smoothing 來變更使用者想要套用的時態性平滑處理量。[Feature] Provide new API k4abt_tracker_set_temporal_smoothing to change the amount of temporal smoothing that the user wants to apply.
  • 特徵新增 c + + 包裝函式 k4abt. 措施。[Feature] Add C++ wrapper k4abt.hpp.
  • 特徵新增版本定義標頭 k4abtversion。[Feature] Add version definition header k4abtversion.h.
  • [Bug 修正]修正導致 CPU 使用量極高的 bug。[Bug Fix] Fix bug that caused extremely high CPU usage.
  • [Bug 修正]修正記錄器損毀 bug。[Bug Fix] Fix logger crashing bug.

v 0.9。1v0.9.1

  • [Bug 修正]修正終結追蹤器時的記憶體遺漏[Bug Fix] Fix memory leak when destroying tracker
  • [Bug 修正]遺失相依性的較佳錯誤訊息[Bug Fix] Better error messages for missing dependencies
  • [Bug 修正]建立第二個追蹤器實例時,失敗而不會損毀[Bug Fix] Fail without crashing when creating a second tracker instance
  • [Bug 修正]記錄器環境變數現在可以正常運作[Bug Fix] Logger environmental variables now work correctly
  • Linux 支援Linux support

v 0.9。0v0.9.0

  • [重大變更]將 SDK 相依性降級至 CUDA 10.0 (從 CUDA 10.1)。[Breaking Change] Downgraded the SDK dependency to CUDA 10.0 (from CUDA 10.1). ONNX 執行時間正式僅支援 CUDA 10.0。ONNX runtime officially only supports up to CUDA 10.0.
  • [重大變更]已切換至 ONNX 執行時間,而不是 Tensorflow 執行時間。[Breaking Change] Switched to ONNX runtime instead of Tensorflow runtime. 減少第一個畫面格啟動時間和記憶體使用量。Reduces the first frame launching time and memory usage. 它也會減少 SDK 二進位檔大小。It also reduces the SDK binary size.
  • [API 變更]已重新命名 k4abt_tracker_queue_capture()k4abt_tracker_enqueue_capture()[API Change] Renamed k4abt_tracker_queue_capture() to k4abt_tracker_enqueue_capture()
  • [API 變更]分解 k4abt_frame_get_body() 成兩個不同的函式: k4abt_frame_get_body_skeleton()k4abt_frame_get_body_id()[API Change] Broke k4abt_frame_get_body() into two separate functions: k4abt_frame_get_body_skeleton() and k4abt_frame_get_body_id(). 現在您可以查詢本文識別碼,而不一定要複製整個基本架構結構。Now you can query the body ID without always copying the whole skeleton structure.
  • [API 變更]已新增函式 k4abt_frame_get_timestamp_usec() ,以簡化使用者查詢主體框架時間戳記的步驟。[API Change] Added k4abt_frame_get_timestamp_usec() function to simplify the steps for the users to query body frame timestamp.
  • 進一步改善本文追蹤演算法的精確度和追蹤可靠性Further improved the body tracking algorithm accuracy and tracking reliability

後續步驟Next steps