Implement windows version to get camera devices in bindings

Author: RHeckerIntel

openvino_bindings/src/BUILD

@@ -5,6 +5,7 @@ cc_library(
     deps = [
         "@nlohmann_json//:json",
         "//third_party:opencv",
+        "//src/utils:input_devices",
         "//src/utils:status",
         "//src/utils:utils",
         "//src/image:image_inference",

openvino_bindings/src/bindings.cc

@@ -20,6 +20,7 @@
 #include "src/vlm/vlm_inference.h"
 #include "src/utils/errors.h"
 #include "src/utils/utils.h"
+#include "src/utils/input_devices.h"
 #include "src/utils/status.h"
 #include "src/image/json_serialization.h"
 #include "src/image/csv_serialization.h"
@@ -501,6 +502,28 @@ StatusOrDevices* getAvailableDevices() {
     return new StatusOrDevices{OkStatus, "", devices, (int)device_ids.size() + 1};
 }
 
+StatusOrInputDevices* getAvailableCameraDevices() {
+    auto cameras = list_camera_devices();
+    InputDevice* devices = new InputDevice[cameras.size()];
+    int i = 0;
+    for (auto camera: cameras) {
+        devices[i] = { (int)camera.first, strdup(camera.second.c_str()) };
+        i++;
+    }
+
+    return new StatusOrInputDevices{OkStatus, "", devices, (int)cameras.size()};
+}
+
+StatusOrString* pdfExtractText(const char* pdf_path) {
+    try {
+        auto output = sentence_extractor::extract_text_from_pdf(pdf_path);
+        return new StatusOrString{OkStatus, "", strdup(output.c_str())};
+    } catch (...) {
+        auto except = handle_exceptions();
+        return new StatusOrString{except->status, except->message};
+    }
+}
+
 Status* handle_exceptions() {
     try {
         throw;

openvino_bindings/src/bindings.h

@@ -36,6 +36,11 @@ typedef struct {
     const char* name;
 } Device;
 
+typedef struct {
+    int id;
+    const char* name;
+} InputDevice;
+
 typedef struct {
   float start_ts;
   float end_ts;
@@ -138,6 +143,13 @@ typedef struct {
     int size;
 } StatusOrDevices;
 
+typedef struct {
+    enum StatusEnum status;
+    const char* message;
+    InputDevice* value;
+    int size;
+} StatusOrInputDevices;
+
 typedef void (*ImageInferenceCallbackFunction)(StatusOrString*);
 typedef void (*LLMInferenceCallbackFunction)(StatusOrString*);
 typedef void (*VLMInferenceCallbackFunction)(StatusOrString*);
@@ -196,6 +208,8 @@ EXPORT StatusOrInt* speechToTextVideoDuration(CSpeechToText instance);
 EXPORT StatusOrWhisperModelResponse* speechToTextTranscribe(CSpeechToText instance, int start, int duration, const char* language);
 
 EXPORT StatusOrDevices* getAvailableDevices();
+EXPORT StatusOrInputDevices* getAvailableCameraDevices();
+
 Status* handle_exceptions();
 
 //extern "C" void report_rss();

openvino_bindings/src/utils/BUILD

@@ -17,6 +17,20 @@ cc_library(
     ],
 )
 
+cc_library(
+    name = "input_devices",
+    hdrs = ["input_devices.h"],
+    deps = [
+         ":errors",
+    ],
+    linkopts = [
+        "mfplat.lib",
+        "mf.lib",
+        "mfuuid.lib",
+        "ole32.lib",
+    ],
+)
+
 cc_library(
     name = "metrics",
     hdrs = [

openvino_bindings/src/utils/input_devices.h

@@ -0,0 +1,86 @@
+#ifndef WINDOWS_INPUT_DEVICES_H_
+#define WINDOWS_INPUT_DEVICES_H_
+
+#include <iostream>
+#include <vector>
+#include <map>
+#include <string>
+
+#include "status.h"
+#include "errors.h"
+
+
+#if _WIN32
+#include <windows.h>
+#include <mfapi.h>
+#include <mfidl.h>
+#include <mfobjects.h>
+#include <mferror.h>
+#pragma comment(lib, "mfplat.lib")
+#pragma comment(lib, "mf.lib")
+
+std::map<size_t, std::string> list_camera_devices() {
+    HRESULT hr = MFStartup(MF_VERSION);
+    if (FAILED(hr)) {
+        throw api_error(InputDeviceError, "MFStartup failed.");
+    }
+
+    IMFAttributes* pAttributes = nullptr;
+    hr = MFCreateAttributes(&pAttributes, 1);
+    if (FAILED(hr)) {
+        MFShutdown();
+        throw api_error(InputDeviceError, "MFCreateAttributes failed.");
+    }
+
+    // Specify that we want video capture devices
+    hr = pAttributes->SetGUID(MF_DEVSOURCE_ATTRIBUTE_SOURCE_TYPE, MF_DEVSOURCE_ATTRIBUTE_SOURCE_TYPE_VIDCAP_GUID);
+    if (FAILED(hr)) {
+        pAttributes->Release();
+        MFShutdown();
+        throw api_error(InputDeviceError, "SetGUID failed.");
+    }
+
+    IMFActivate** ppDevices = nullptr;
+    UINT32 count = 0;
+
+
+    std::map<size_t, std::string> cameras = {};
+    hr = MFEnumDeviceSources(pAttributes, &ppDevices, &count);
+    if (SUCCEEDED(hr)) {
+        for (UINT32 i = 0; i < count; i++) {
+            WCHAR* szFriendlyName = nullptr;
+            UINT32 cchName = 0;
+            hr = ppDevices[i]->GetAllocatedString(MF_DEVSOURCE_ATTRIBUTE_FRIENDLY_NAME, &szFriendlyName, &cchName);
+            if (SUCCEEDED(hr)) {
+                std::wstring ws(szFriendlyName);
+                cameras.insert({i, std::string(ws.begin(), ws.end())});
+                //std::wcout << L"[" << i << L"]: " << szFriendlyName << std::endl;
+                CoTaskMemFree(szFriendlyName);
+            }
+            ppDevices[i]->Release();
+        }
+        CoTaskMemFree(ppDevices);
+    } else {
+        std::cerr << "No camera devices found.\n";
+    }
+
+    pAttributes->Release();
+    MFShutdown();
+
+    return cameras;
+}
+#elif __APPLE__
+std::map<size_t, std::string> list_camera_devices() {
+    return {};
+}
+#elif __linux__
+std::map<size_t, std::string> list_camera_devices() {
+    return {};
+}
+#else
+std::map<size_t, std::string> list_camera_devices() {
+    throw api_error(InputDeviceError, "Unsupported platform.");
+}
+#endif
+
+#endif // WINDOWS_INPUT_DEVICES_H_

openvino_bindings/src/utils/status.h

@@ -35,6 +35,8 @@ enum StatusEnum {
   SpeechToTextFileNotOpened = -81,
   SpeechToTextChunkHasNoData = -82,
   SpeechToTextChunkOutOfBounds = -83,
+
+  InputDeviceError = -90,
 };
 
 #endif // STATUS_H_