blob: 4764d3123c3d70faf10265cf1cd85a1659aee717 [file] [log] [blame]
Avi Drissman8ba1bad2022-09-13 19:22:361// Copyright 2022 The Chromium Authors
Ramin Halavati26dcce22022-02-23 13:11:142// Use of this source code is governed by a BSD-style license that can be
3// found in the LICENSE file.
4
Abigail Klein7a63c572024-02-28 20:45:095#include "services/screen_ai/screen_ai_service_impl.h"
Ramin Halavati26dcce22022-02-23 13:11:146
Ramin Halavati55fdf4d2022-11-07 06:35:187#include <memory>
Ramin Halavati63d9c252024-07-29 06:59:178#include <optional>
Ramin Halavatifd712d8e2022-10-19 05:11:279#include <utility>
Ramin Halavatib044504d2022-10-24 06:36:3410#include <vector>
Ramin Halavatifd712d8e2022-10-19 05:11:2711
12#include "base/check.h"
Ramin Halavatiad969ef2024-05-21 06:59:4713#include "base/check_is_test.h"
Tom Sepeze9b4dd552024-08-14 22:44:2914#include "base/compiler_specific.h"
Ramin Halavati55fdf4d2022-11-07 06:35:1815#include "base/functional/bind.h"
16#include "base/location.h"
Ramin Halavatifafff152022-12-05 09:39:3817#include "base/logging.h"
Kyungjun Lee9d755a02022-11-08 17:57:1118#include "base/metrics/histogram_functions.h"
Ramin Halavati5ff855c2022-04-27 16:30:4819#include "base/process/process.h"
Sean Maher5b9af51f2022-11-21 15:32:4720#include "base/task/single_thread_task_runner.h"
Ramin Halavati55fdf4d2022-11-07 06:35:1821#include "base/task/thread_pool.h"
Abigail Klein7a63c572024-02-28 20:45:0922#include "services/screen_ai/buildflags/buildflags.h"
Ramin Halavati4c47eb82024-12-03 19:06:4223#include "services/screen_ai/proto/chrome_screen_ai.pb.h"
Abigail Klein7a63c572024-02-28 20:45:0924#include "services/screen_ai/proto/main_content_extractor_proto_convertor.h"
25#include "services/screen_ai/proto/visual_annotator_proto_convertor.h"
Ramin Halavati4c47eb82024-12-03 19:06:4226#include "services/screen_ai/public/cpp/metrics.h"
Abigail Klein7a63c572024-02-28 20:45:0927#include "services/screen_ai/public/cpp/utilities.h"
Ramin Halavatieddadb62022-05-04 17:29:4928#include "ui/accessibility/accessibility_features.h"
Ramin Halavati4f904d0a2024-06-25 15:32:3229#include "ui/accessibility/ax_node.h"
30#include "ui/accessibility/ax_tree.h"
Ramin Halavati84231072022-08-17 08:02:2031#include "ui/accessibility/ax_tree_id.h"
Nektarios Paisiosf73d6972022-06-04 11:32:2532#include "ui/gfx/geometry/rect_f.h"
Ramin Halavati367352a2022-04-14 06:00:2333
Ramin Halavatia9b50a102024-02-07 18:37:2334#if BUILDFLAG(USE_FAKE_SCREEN_AI)
Abigail Klein7a63c572024-02-28 20:45:0935#include "services/screen_ai/screen_ai_library_wrapper_fake.h"
Ramin Halavati767b8fc2024-02-02 06:23:0836#else
Abigail Klein7a63c572024-02-28 20:45:0937#include "services/screen_ai/screen_ai_library_wrapper_impl.h"
Ramin Halavati767b8fc2024-02-02 06:23:0838#endif
39
Ramin Halavati69bf7752022-04-04 09:58:1540namespace screen_ai {
Ramin Halavati26dcce22022-02-23 13:11:1441
Ramin Halavati6d4fc2b2022-06-16 15:33:3542namespace {
43
Ramin Halavati2acf2362024-08-02 14:22:5344// How often it would be checked that the service is idle and can be shutdown.
Ramin Halavatidc0b2e512024-08-16 17:39:0945constexpr base::TimeDelta kIdleCheckingDelay = base::Minutes(5);
Ramin Halavati2acf2362024-08-02 14:22:5346
Ramin Halavati6ff56412024-08-13 07:15:5447// How long after all clients are disconnected, it is checked if service is
48// idle.
Ramin Halavati2fa45b82024-08-15 16:46:5349constexpr base::TimeDelta kCoolDownTime = base::Seconds(10);
Ramin Halavati6ff56412024-08-13 07:15:5450
Ramin Halavatiad969ef2024-05-21 06:59:4751// These values are persisted to logs. Entries should not be renumbered and
52// numeric values should never be reused.
Ramin Halavati29cc1152024-10-27 17:30:2853// See `screen_ai_service.mojom` for more info.
Ramin Halavatiad969ef2024-05-21 06:59:4754enum class OcrClientTypeForMetrics {
55 kTest = 0,
56 kPdfViewer = 1,
57 kLocalSearch = 2,
58 kCameraApp = 3,
Ramin Halavati29cc1152024-10-27 17:30:2859 kNotUsed = 4, // Can be used for a new client.
Ramin Halavatiad969ef2024-05-21 06:59:4760 kMediaApp = 5,
Michelle Chen5192d522024-10-22 23:38:2261 kScreenshotTextDetection,
62 kMaxValue = kScreenshotTextDetection
Ramin Halavatiad969ef2024-05-21 06:59:4763};
64
65OcrClientTypeForMetrics GetClientType(mojom::OcrClientType client_type) {
66 switch (client_type) {
67 case mojom::OcrClientType::kTest:
68 CHECK_IS_TEST();
69 return OcrClientTypeForMetrics::kTest;
70 case mojom::OcrClientType::kPdfViewer:
71 return OcrClientTypeForMetrics::kPdfViewer;
72 case mojom::OcrClientType::kLocalSearch:
73 return OcrClientTypeForMetrics::kLocalSearch;
74 case mojom::OcrClientType::kCameraApp:
75 return OcrClientTypeForMetrics::kCameraApp;
Ramin Halavatiad969ef2024-05-21 06:59:4776 case mojom::OcrClientType::kMediaApp:
77 return OcrClientTypeForMetrics::kMediaApp;
Michelle Chen5192d522024-10-22 23:38:2278 case mojom::OcrClientType::kScreenshotTextDetection:
79 return OcrClientTypeForMetrics::kScreenshotTextDetection;
Ramin Halavatiad969ef2024-05-21 06:59:4780 }
81}
82
Ramin Halavati89a36c92023-05-05 15:01:0383ui::AXTreeUpdate ConvertVisualAnnotationToTreeUpdate(
Ramin Halavatidcb09a172024-07-15 08:28:5584 std::optional<chrome_screen_ai::VisualAnnotation>& annotation_proto,
Ramin Halavati89a36c92023-05-05 15:01:0385 const gfx::Rect& image_rect) {
86 if (!annotation_proto) {
87 VLOG(0) << "Screen AI library could not process snapshot or no OCR data.";
88 return ui::AXTreeUpdate();
89 }
90
91 return VisualAnnotationToAXTreeUpdate(*annotation_proto, image_rect);
92}
93
Abigail Klein18673702024-03-05 20:59:0694ui::AXNodeID ComputeMainNode(
95 const ui::AXTree* tree,
96 const std::vector<ui::AXNodeID>& content_node_ids) {
97 ui::AXNode* front = tree->GetFromId(content_node_ids.front());
98 ui::AXNode* back = tree->GetFromId(content_node_ids.back());
99 ui::AXNode* main = front->GetLowestCommonAncestor(*back);
100 return main->id();
101}
102
Ramin Halavati26ae6b72022-11-10 06:45:05103} // namespace
104
Ramin Halavati39bab0d2024-01-30 06:05:01105// The library accepts simple pointers to model data retrieval functions, hence
106// callback functions with linked object are not safe to pass.
Ramin Halavati63d9c252024-07-29 06:59:17107// This global variable keeps the pointer the only instance of this class.
108ModelDataHolder* g_model_data_holder_instance = nullptr;
Ramin Halavati8ced6392023-09-25 07:27:50109
Ramin Halavati63d9c252024-07-29 06:59:17110// Keeps the handles of model files, and replies to calls for copying their
111// content.
112class ModelDataHolder {
Ramin Halavati8ced6392023-09-25 07:27:50113 public:
Ramin Halavati63d9c252024-07-29 06:59:17114 ModelDataHolder() {
115 CHECK(!g_model_data_holder_instance);
116 g_model_data_holder_instance = this;
117 }
Ramin Halavati8ced6392023-09-25 07:27:50118
Ramin Halavati63d9c252024-07-29 06:59:17119 ModelDataHolder(const ModelDataHolder&) = delete;
120 ModelDataHolder& operator=(const ModelDataHolder&) = delete;
121
122 ~ModelDataHolder() {
123 CHECK_EQ(g_model_data_holder_instance, this);
124 g_model_data_holder_instance = nullptr;
Ramin Halavati8ced6392023-09-25 07:27:50125 }
126
127 // Returns 0 if file is not found.
128 static uint32_t GetDataSize(const char* relative_file_path) {
Ramin Halavati63d9c252024-07-29 06:59:17129 CHECK(g_model_data_holder_instance);
130 base::File* model_file =
131 g_model_data_holder_instance->GetModelFile(relative_file_path);
132 return model_file ? model_file->GetLength() : 0;
Ramin Halavati8ced6392023-09-25 07:27:50133 }
134
Ramin Halavati63d9c252024-07-29 06:59:17135 // Copies content of the file in `relative_file_path` to `buffer`. Expects
136 // that `buffer_size` would be enough for the entire file content.
Ramin Halavati8ced6392023-09-25 07:27:50137 static void CopyData(const char* relative_file_path,
138 uint32_t buffer_size,
139 char* buffer) {
Ramin Halavati63d9c252024-07-29 06:59:17140 CHECK(g_model_data_holder_instance);
141 base::File* model_file =
142 g_model_data_holder_instance->GetModelFile(relative_file_path);
143 CHECK(model_file);
144
145 int64_t length = model_file->GetLength();
146 CHECK_GE(buffer_size, length);
Tom Sepeze9b4dd552024-08-14 22:44:29147 CHECK_EQ(UNSAFE_TODO(model_file->Read(0, buffer, length)), length);
Ramin Halavati8ced6392023-09-25 07:27:50148 }
149
Ramin Halavati63d9c252024-07-29 06:59:17150 void AddModelFiles(base::flat_map<base::FilePath, base::File> model_files) {
151 for (auto& model_file : model_files) {
152 model_files_[model_file.first.MaybeAsASCII()] =
153 std::move(model_file.second);
Ramin Halavati39bab0d2024-01-30 06:05:01154 }
155 }
Ramin Halavati8ced6392023-09-25 07:27:50156
Ramin Halavati63d9c252024-07-29 06:59:17157 // Returns the file handle for `relative_file_path` if it exists.
158 base::File* GetModelFile(const char* relative_file_path) {
159 if (!base::Contains(model_files_, relative_file_path)) {
160 return nullptr;
161 }
162 return &model_files_[relative_file_path];
163 }
164
Ramin Halavati8ced6392023-09-25 07:27:50165 private:
Ramin Halavati63d9c252024-07-29 06:59:17166 std::map<std::string, base::File> model_files_;
Ramin Halavati8ced6392023-09-25 07:27:50167};
168
Ramin Halavati26ae6b72022-11-10 06:45:05169ScreenAIService::ScreenAIService(
Ramin Halavatic1e4fa92023-05-17 17:22:06170 mojo::PendingReceiver<mojom::ScreenAIServiceFactory> receiver)
171 : factory_receiver_(this, std::move(receiver)),
172 ocr_receiver_(this),
Ramin Halavatiad969ef2024-05-21 06:59:47173 main_content_extraction_receiver_(this) {
Ramin Halavati6ff56412024-08-13 07:15:54174 screen2x_main_content_extractors_.set_disconnect_handler(
175 base::BindRepeating(&ScreenAIService::CheckIdleStateAfterDelay,
176 weak_ptr_factory_.GetWeakPtr()));
177 screen_ai_annotators_.set_disconnect_handler(
178 base::BindRepeating(&ScreenAIService::OcrReceiverDisconnected,
179 weak_ptr_factory_.GetWeakPtr()));
Ramin Halavati63d9c252024-07-29 06:59:17180 model_data_holder_ = std::make_unique<ModelDataHolder>();
Ramin Halavati2acf2362024-08-02 14:22:53181 idle_checking_timer_ = std::make_unique<base::RepeatingTimer>();
182 idle_checking_timer_->Start(FROM_HERE, kIdleCheckingDelay, this,
183 &ScreenAIService::ShutDownIfNoClients);
Ramin Halavatiad969ef2024-05-21 06:59:47184}
Ramin Halavati26ae6b72022-11-10 06:45:05185
186ScreenAIService::~ScreenAIService() = default;
187
Ramin Halavati8ced6392023-09-25 07:27:50188void ScreenAIService::LoadLibrary(const base::FilePath& library_path) {
Abigail Klein7a63c572024-02-28 20:45:09189 // The ScopedBlockingCall in LoadLibrary guarantees that this is not run on
190 // the UI thread.
Ramin Halavatia9b50a102024-02-07 18:37:23191#if BUILDFLAG(USE_FAKE_SCREEN_AI)
Ramin Halavati767b8fc2024-02-02 06:23:08192 library_ = std::make_unique<ScreenAILibraryWrapperFake>();
193#else
194 library_ = std::make_unique<ScreenAILibraryWrapperImpl>();
195#endif
Ramin Halavati8ced6392023-09-25 07:27:50196
197 bool load_sucessful = library_->Load(library_path);
198 base::UmaHistogramBoolean("Accessibility.ScreenAI.Library.Initialized",
199 load_sucessful);
200
201 if (!load_sucessful) {
202 library_.reset();
203 return;
204 }
205
206 uint32_t version_major;
207 uint32_t version_minor;
208 library_->GetLibraryVersion(version_major, version_minor);
209 VLOG(2) << "Screen AI library version: " << version_major << "."
210 << version_minor;
211
212#if BUILDFLAG(IS_CHROMEOS_ASH)
213 library_->SetLogger();
214#endif
215
216 if (features::IsScreenAIDebugModeEnabled()) {
217 library_->EnableDebugMode();
218 }
219
Ramin Halavati63d9c252024-07-29 06:59:17220 library_->SetFileContentFunctions(&ModelDataHolder::GetDataSize,
221 &ModelDataHolder::CopyData);
Ramin Halavati8ced6392023-09-25 07:27:50222}
223
Ramin Halavatic1e4fa92023-05-17 17:22:06224void ScreenAIService::InitializeMainContentExtraction(
Ramin Halavatid139a0f2023-02-28 13:25:28225 const base::FilePath& library_path,
Ramin Halavatifbc62022023-11-10 11:05:07226 base::flat_map<base::FilePath, base::File> model_files,
Ramin Halavatic1e4fa92023-05-17 17:22:06227 mojo::PendingReceiver<mojom::MainContentExtractionService>
228 main_content_extractor_service_receiver,
229 InitializeMainContentExtractionCallback callback) {
230 if (!library_) {
Ramin Halavati8ced6392023-09-25 07:27:50231 LoadLibrary(library_path);
Ramin Halavatic1e4fa92023-05-17 17:22:06232 }
233
234 if (!library_) {
235 std::move(callback).Run(false);
236 base::Process::TerminateCurrentProcessImmediately(-1);
237 }
238
Ramin Halavati63d9c252024-07-29 06:59:17239 model_data_holder_->AddModelFiles(std::move(model_files));
Ramin Halavatib044504d2022-10-24 06:36:34240
Ramin Halavati8ced6392023-09-25 07:27:50241 bool init_successful = library_->InitMainContentExtraction();
Ramin Halavati1d57c2d2023-05-24 05:18:24242 base::UmaHistogramBoolean(
243 "Accessibility.ScreenAI.MainContentExtraction.Initialized",
244 init_successful);
245 if (!init_successful) {
Ramin Halavatic1e4fa92023-05-17 17:22:06246 std::move(callback).Run(false);
Ramin Halavati89a36c92023-05-05 15:01:03247 return;
Ramin Halavatid139a0f2023-02-28 13:25:28248 }
Ramin Halavati89a36c92023-05-05 15:01:03249
Ramin Halavatic1e4fa92023-05-17 17:22:06250 // This interface should be created only once.
251 CHECK(!main_content_extraction_receiver_.is_bound());
252
253 main_content_extraction_receiver_.Bind(
254 std::move(main_content_extractor_service_receiver));
255
256 std::move(callback).Run(true);
Ramin Halavati2acf2362024-08-02 14:22:53257 main_content_extraction_last_used_ = base::TimeTicks::Now();
Ramin Halavatic1e4fa92023-05-17 17:22:06258}
259
260void ScreenAIService::InitializeOCR(
261 const base::FilePath& library_path,
Ramin Halavatifbc62022023-11-10 11:05:07262 base::flat_map<base::FilePath, base::File> model_files,
Ramin Halavatic1e4fa92023-05-17 17:22:06263 mojo::PendingReceiver<mojom::OCRService> ocr_service_receiver,
264 InitializeOCRCallback callback) {
265 if (!library_) {
Ramin Halavati8ced6392023-09-25 07:27:50266 LoadLibrary(library_path);
Ramin Halavatic1e4fa92023-05-17 17:22:06267 }
268
269 if (!library_) {
270 std::move(callback).Run(false);
271 base::Process::TerminateCurrentProcessImmediately(-1);
272 }
273
Ramin Halavati63d9c252024-07-29 06:59:17274 model_data_holder_->AddModelFiles(std::move(model_files));
275
Ramin Halavati0e2ae5742023-11-10 05:57:54276 bool init_successful = library_->InitOCR();
Ramin Halavati35d129f2023-06-22 16:37:09277 base::UmaHistogramBoolean("Accessibility.ScreenAI.OCR.Initialized",
Ramin Halavati1d57c2d2023-05-24 05:18:24278 init_successful);
Ramin Halavati480584332023-08-02 06:38:40279
Ramin Halavati1d57c2d2023-05-24 05:18:24280 if (!init_successful) {
Ramin Halavatic1e4fa92023-05-17 17:22:06281 std::move(callback).Run(false);
282 return;
283 }
284
285 // This interface should be created only once.
286 CHECK(!ocr_receiver_.is_bound());
287
288 ocr_receiver_.Bind(std::move(ocr_service_receiver));
289
290 std::move(callback).Run(true);
Ramin Halavati2acf2362024-08-02 14:22:53291 ocr_last_used_ = base::TimeTicks::Now();
Ramin Halavatib044504d2022-10-24 06:36:34292}
Ramin Halavati26dcce22022-02-23 13:11:14293
294void ScreenAIService::BindAnnotator(
295 mojo::PendingReceiver<mojom::ScreenAIAnnotator> annotator) {
296 screen_ai_annotators_.Add(this, std::move(annotator));
297}
298
Ramin Halavatieddadb62022-05-04 17:29:49299void ScreenAIService::BindMainContentExtractor(
300 mojo::PendingReceiver<mojom::Screen2xMainContentExtractor>
301 main_content_extractor) {
Ramin Halavati6ff56412024-08-13 07:15:54302 screen2x_main_content_extractors_.Add(this,
303 std::move(main_content_extractor));
Ramin Halavatieddadb62022-05-04 17:29:49304}
305
Arthur Sonzognic571efb2024-01-26 20:26:18306std::optional<chrome_screen_ai::VisualAnnotation>
Ramin Halavati7716b032024-08-19 14:56:20307ScreenAIService::PerformOcrAndRecordMetrics(const SkBitmap& image) {
308 CHECK(base::Contains(ocr_client_types_,
309 screen_ai_annotators_.current_receiver()));
310 mojom::OcrClientType client_type =
311 ocr_client_types_.find(screen_ai_annotators_.current_receiver())->second;
Ramin Halavatiad969ef2024-05-21 06:59:47312 base::UmaHistogramEnumeration("Accessibility.ScreenAI.OCR.ClientType",
Ramin Halavati7716b032024-08-19 14:56:20313 GetClientType(client_type));
Ramin Halavatiad969ef2024-05-21 06:59:47314
Ramin Halavati2acf2362024-08-02 14:22:53315 ocr_last_used_ = base::TimeTicks::Now();
Ramin Halavati7252cd02023-05-10 07:21:33316 auto result = library_->PerformOcr(image);
Ramin Halavati2acf2362024-08-02 14:22:53317 base::TimeDelta elapsed_time = base::TimeTicks::Now() - ocr_last_used_;
Ramin Halavati2650ebe2023-11-27 19:56:35318 int lines_count = result ? result->lines_size() : 0;
Ramin Halavati3ae33ca2024-01-01 12:58:39319 unsigned image_size = image.width() * image.height();
Ramin Halavati2650ebe2023-11-27 19:56:35320 VLOG(1) << "OCR returned " << lines_count << " lines in " << elapsed_time;
Ramin Halavati7252cd02023-05-10 07:21:33321
Ramin Halavati18aba872024-08-23 14:33:05322 if (!result) {
323 base::UmaHistogramEnumeration(
324 "Accessibility.ScreenAI.OCR.Failed.ClientType",
325 GetClientType(client_type));
326 }
Ramin Halavati81b6d922024-06-07 14:20:34327 base::UmaHistogramBoolean("Accessibility.ScreenAI.OCR.Successful",
328 result.has_value());
Ramin Halavati2650ebe2023-11-27 19:56:35329 base::UmaHistogramCounts100("Accessibility.ScreenAI.OCR.LinesCount",
330 lines_count);
Ramin Halavatia0548b262023-06-27 05:56:47331 base::UmaHistogramCounts10M("Accessibility.ScreenAI.OCR.ImageSize10M",
Ramin Halavati3ae33ca2024-01-01 12:58:39332 image_size);
333 if (image_size < 500 * 500) {
334 base::UmaHistogramTimes("Accessibility.ScreenAI.OCR.Latency.Small",
335 elapsed_time);
336 } else if (image_size < 1000 * 1000) {
337 base::UmaHistogramTimes("Accessibility.ScreenAI.OCR.Latency.Medium",
338 elapsed_time);
339 } else if (image_size < 2000 * 2000) {
340 base::UmaHistogramTimes("Accessibility.ScreenAI.OCR.Latency.Large",
341 elapsed_time);
342 } else {
343 base::UmaHistogramTimes("Accessibility.ScreenAI.OCR.Latency.XLarge",
344 elapsed_time);
345 }
Ramin Halavati2650ebe2023-11-27 19:56:35346
Ramin Halavati7716b032024-08-19 14:56:20347 // MediaApp provides OCR for ChromeOS PDF viewer.
348 if (client_type == mojom::OcrClientType::kPdfViewer ||
349 client_type == mojom::OcrClientType::kMediaApp) {
Ramin Halavati2650ebe2023-11-27 19:56:35350 base::UmaHistogramCounts100("Accessibility.ScreenAI.OCR.LinesCount.PDF",
351 lines_count);
352 base::UmaHistogramTimes("Accessibility.ScreenAI.OCR.Time.PDF",
353 elapsed_time);
354 base::UmaHistogramCounts10M("Accessibility.ScreenAI.OCR.ImageSize.PDF",
355 image.width() * image.height());
Ramin Halavati4c47eb82024-12-03 19:06:42356
357 std::optional<uint64_t> most_detected_language =
358 GetMostDetectedLanguageInOcrData(*result);
359 if (most_detected_language.has_value()) {
360 base::UmaHistogramSparse(
361 "Accessibility.ScreenAI.OCR.MostDetectedLanguage.PDF",
362 most_detected_language.value());
363 }
Ramin Halavati2650ebe2023-11-27 19:56:35364 }
365
Ramin Halavati7252cd02023-05-10 07:21:33366 return result;
367}
368
Ramin Halavatiad969ef2024-05-21 06:59:47369void ScreenAIService::SetClientType(mojom::OcrClientType client_type) {
370 ocr_client_types_[screen_ai_annotators_.current_receiver()] = client_type;
371}
372
Dmitry Grebenyukc49b95ee2023-05-02 05:52:07373void ScreenAIService::PerformOcrAndReturnAnnotation(
374 const SkBitmap& image,
375 PerformOcrAndReturnAnnotationCallback callback) {
Arthur Sonzognic571efb2024-01-26 20:26:18376 std::optional<chrome_screen_ai::VisualAnnotation> annotation_proto =
Ramin Halavati7716b032024-08-19 14:56:20377 PerformOcrAndRecordMetrics(image);
Dmitry Grebenyukc49b95ee2023-05-02 05:52:07378
Ramin Halavati89a36c92023-05-05 15:01:03379 if (annotation_proto) {
380 std::move(callback).Run(ConvertProtoToVisualAnnotation(*annotation_proto));
381 return;
Dmitry Grebenyukc49b95ee2023-05-02 05:52:07382 }
Ramin Halavati89a36c92023-05-05 15:01:03383
384 std::move(callback).Run(mojom::VisualAnnotation::New());
Dmitry Grebenyukc49b95ee2023-05-02 05:52:07385}
386
Kyungjun Lee90461522023-04-25 06:50:09387void ScreenAIService::PerformOcrAndReturnAXTreeUpdate(
Ramin Halavatib446c022023-03-29 15:36:36388 const SkBitmap& image,
Kyungjun Lee90461522023-04-25 06:50:09389 PerformOcrAndReturnAXTreeUpdateCallback callback) {
Arthur Sonzognic571efb2024-01-26 20:26:18390 std::optional<chrome_screen_ai::VisualAnnotation> annotation_proto =
Ramin Halavati7716b032024-08-19 14:56:20391 PerformOcrAndRecordMetrics(image);
Ramin Halavati89a36c92023-05-05 15:01:03392 ui::AXTreeUpdate update = ConvertVisualAnnotationToTreeUpdate(
393 annotation_proto, gfx::Rect(image.width(), image.height()));
Kyungjun Lee90461522023-04-25 06:50:09394
Ramin Halavati89a36c92023-05-05 15:01:03395 // The original caller is always replied to, and an empty AXTreeUpdate tells
396 // that the annotation function was not successful.
397 std::move(callback).Run(update);
Ramin Halavatib446c022023-03-29 15:36:36398}
399
400void ScreenAIService::ExtractMainContent(const ui::AXTreeUpdate& snapshot,
Ramin Halavatib446c022023-03-29 15:36:36401 ExtractMainContentCallback callback) {
Ramin Halavati2acf2362024-08-02 14:22:53402 main_content_extraction_last_used_ = base::TimeTicks::Now();
Abigail Klein18673702024-03-05 20:59:06403 ui::AXTree tree;
404 std::optional<std::vector<int32_t>> content_node_ids;
405 bool success = ExtractMainContentInternal(snapshot, tree, content_node_ids);
Abigail Klein18673702024-03-05 20:59:06406
407 if (success) {
408 std::move(callback).Run(*content_node_ids);
409 } else {
410 std::move(callback).Run(std::vector<int32_t>());
411 }
412}
413
414void ScreenAIService::ExtractMainNode(const ui::AXTreeUpdate& snapshot,
415 ExtractMainNodeCallback callback) {
416 ui::AXTree tree;
417 std::optional<std::vector<int32_t>> content_node_ids;
418 bool success = ExtractMainContentInternal(snapshot, tree, content_node_ids);
419
420 if (success) {
421 ui::AXNodeID main_node_id = ComputeMainNode(&tree, *content_node_ids);
422 std::move(callback).Run(main_node_id);
423 } else {
424 std::move(callback).Run(ui::kInvalidAXNodeID);
425 }
426}
427
428bool ScreenAIService::ExtractMainContentInternal(
429 const ui::AXTreeUpdate& snapshot,
430 ui::AXTree& tree,
431 std::optional<std::vector<int32_t>>& content_node_ids) {
Ramin Halavatib446c022023-03-29 15:36:36432 // Early return if input is empty.
433 if (snapshot.nodes.empty()) {
Abigail Klein18673702024-03-05 20:59:06434 return false;
Ramin Halavatib33fc0c2022-05-06 09:32:22435 }
Ramin Halavatie2549a92022-08-02 07:43:19436
Abigail Klein18673702024-03-05 20:59:06437 // Deserialize the snapshot and reserialize it to a view hierarchy proto.
438 CHECK(tree.Unserialize(snapshot));
Ramin Halavati62e034d2024-08-02 08:58:22439 std::optional<ViewHierarchyAndTreeSize> converted_snapshot =
440 SnapshotToViewHierarchy(tree);
441 if (!converted_snapshot) {
442 VLOG(0) << "Proto not generated.";
443 return false;
444 }
445
Ramin Halavati62e034d2024-08-02 08:58:22446 content_node_ids =
447 library_->ExtractMainContent(converted_snapshot->serialized_proto);
Ramin Halavati81b6d922024-06-07 14:20:34448 base::UmaHistogramBoolean(
449 "Accessibility.ScreenAI.MainContentExtraction.Successful",
450 content_node_ids.has_value());
Abigail Klein18673702024-03-05 20:59:06451 if (content_node_ids.has_value() && content_node_ids->size() > 0) {
Ramin Halavati5aad701d2023-05-09 17:23:29452 VLOG(2) << "Screen2x returned " << content_node_ids->size() << " node ids.";
Abigail Klein18673702024-03-05 20:59:06453 return true;
454 } else {
455 VLOG(0) << "Screen2x returned no results.";
456 return false;
Ramin Halavati5aad701d2023-05-09 17:23:29457 }
Abigail Klein18673702024-03-05 20:59:06458}
Ramin Halavati89a36c92023-05-05 15:01:03459
Abigail Klein18673702024-03-05 20:59:06460ui::AXNodeID ScreenAIService::ComputeMainNodeForTesting(
461 const ui::AXTree* tree,
462 const std::vector<ui::AXNodeID>& content_node_ids) {
463 return ComputeMainNode(tree, content_node_ids);
Abigail Kleindfdde352023-01-27 21:03:10464}
465
Ramin Halavati6ff56412024-08-13 07:15:54466void ScreenAIService::OcrReceiverDisconnected() {
Ramin Halavatiad969ef2024-05-21 06:59:47467 auto entry = ocr_client_types_.find(screen_ai_annotators_.current_receiver());
468 if (entry != ocr_client_types_.end()) {
469 ocr_client_types_.erase(entry);
470 }
Ramin Halavati6ff56412024-08-13 07:15:54471
472 CheckIdleStateAfterDelay();
473}
474
475void ScreenAIService::CheckIdleStateAfterDelay() {
476 // Check if service is idle, a little after the client disconnects.
477 base::SequencedTaskRunner::GetCurrentDefault()->PostDelayedTask(
478 FROM_HERE,
479 base::BindOnce(&ScreenAIService::ShutDownIfNoClients,
480 weak_ptr_factory_.GetWeakPtr()),
481 kCoolDownTime);
Ramin Halavatiad969ef2024-05-21 06:59:47482}
483
Ramin Halavati3842e7e2024-07-31 04:39:21484void ScreenAIService::ShutDownIfNoClients() {
Ramin Halavati2acf2362024-08-02 14:22:53485 const base::TimeTicks kIdlenessThreshold =
486 base::TimeTicks::Now() - kIdleCheckingDelay;
Ramin Halavatidc0b2e512024-08-16 17:39:09487 bool ocr_not_needed =
488 !screen_ai_annotators_.size() || ocr_last_used_ < kIdlenessThreshold;
489 bool main_content_extractioncan_not_needed =
490 !screen2x_main_content_extractors_.size() ||
491 main_content_extraction_last_used_ < kIdlenessThreshold;
492
493 if (ocr_not_needed && main_content_extractioncan_not_needed) {
494 VLOG(2) << "Shutting down since no client or idle.";
495 base::Process::TerminateCurrentProcessImmediately(0);
Ramin Halavati2acf2362024-08-02 14:22:53496 }
Ramin Halavati3842e7e2024-07-31 04:39:21497}
498
Ramin Halavati26dcce22022-02-23 13:11:14499} // namespace screen_ai