Avi Drissman | 8ba1bad | 2022-09-13 19:22:36 | [diff] [blame] | 1 | // Copyright 2022 The Chromium Authors |
Ramin Halavati | 26dcce2 | 2022-02-23 13:11:14 | [diff] [blame] | 2 | // Use of this source code is governed by a BSD-style license that can be |
| 3 | // found in the LICENSE file. |
| 4 | |
Abigail Klein | 7a63c57 | 2024-02-28 20:45:09 | [diff] [blame] | 5 | #include "services/screen_ai/screen_ai_service_impl.h" |
Ramin Halavati | 26dcce2 | 2022-02-23 13:11:14 | [diff] [blame] | 6 | |
Ramin Halavati | 55fdf4d | 2022-11-07 06:35:18 | [diff] [blame] | 7 | #include <memory> |
Ramin Halavati | 63d9c25 | 2024-07-29 06:59:17 | [diff] [blame] | 8 | #include <optional> |
Ramin Halavati | fd712d8e | 2022-10-19 05:11:27 | [diff] [blame] | 9 | #include <utility> |
Ramin Halavati | b044504d | 2022-10-24 06:36:34 | [diff] [blame] | 10 | #include <vector> |
Ramin Halavati | fd712d8e | 2022-10-19 05:11:27 | [diff] [blame] | 11 | |
| 12 | #include "base/check.h" |
Ramin Halavati | ad969ef | 2024-05-21 06:59:47 | [diff] [blame] | 13 | #include "base/check_is_test.h" |
Tom Sepez | e9b4dd55 | 2024-08-14 22:44:29 | [diff] [blame] | 14 | #include "base/compiler_specific.h" |
Ramin Halavati | 55fdf4d | 2022-11-07 06:35:18 | [diff] [blame] | 15 | #include "base/functional/bind.h" |
| 16 | #include "base/location.h" |
Ramin Halavati | fafff15 | 2022-12-05 09:39:38 | [diff] [blame] | 17 | #include "base/logging.h" |
Kyungjun Lee | 9d755a0 | 2022-11-08 17:57:11 | [diff] [blame] | 18 | #include "base/metrics/histogram_functions.h" |
Ramin Halavati | 5ff855c | 2022-04-27 16:30:48 | [diff] [blame] | 19 | #include "base/process/process.h" |
Sean Maher | 5b9af51f | 2022-11-21 15:32:47 | [diff] [blame] | 20 | #include "base/task/single_thread_task_runner.h" |
Ramin Halavati | 55fdf4d | 2022-11-07 06:35:18 | [diff] [blame] | 21 | #include "base/task/thread_pool.h" |
Abigail Klein | 7a63c57 | 2024-02-28 20:45:09 | [diff] [blame] | 22 | #include "services/screen_ai/buildflags/buildflags.h" |
Ramin Halavati | 4c47eb8 | 2024-12-03 19:06:42 | [diff] [blame^] | 23 | #include "services/screen_ai/proto/chrome_screen_ai.pb.h" |
Abigail Klein | 7a63c57 | 2024-02-28 20:45:09 | [diff] [blame] | 24 | #include "services/screen_ai/proto/main_content_extractor_proto_convertor.h" |
| 25 | #include "services/screen_ai/proto/visual_annotator_proto_convertor.h" |
Ramin Halavati | 4c47eb8 | 2024-12-03 19:06:42 | [diff] [blame^] | 26 | #include "services/screen_ai/public/cpp/metrics.h" |
Abigail Klein | 7a63c57 | 2024-02-28 20:45:09 | [diff] [blame] | 27 | #include "services/screen_ai/public/cpp/utilities.h" |
Ramin Halavati | eddadb6 | 2022-05-04 17:29:49 | [diff] [blame] | 28 | #include "ui/accessibility/accessibility_features.h" |
Ramin Halavati | 4f904d0a | 2024-06-25 15:32:32 | [diff] [blame] | 29 | #include "ui/accessibility/ax_node.h" |
| 30 | #include "ui/accessibility/ax_tree.h" |
Ramin Halavati | 8423107 | 2022-08-17 08:02:20 | [diff] [blame] | 31 | #include "ui/accessibility/ax_tree_id.h" |
Nektarios Paisios | f73d697 | 2022-06-04 11:32:25 | [diff] [blame] | 32 | #include "ui/gfx/geometry/rect_f.h" |
Ramin Halavati | 367352a | 2022-04-14 06:00:23 | [diff] [blame] | 33 | |
Ramin Halavati | a9b50a10 | 2024-02-07 18:37:23 | [diff] [blame] | 34 | #if BUILDFLAG(USE_FAKE_SCREEN_AI) |
Abigail Klein | 7a63c57 | 2024-02-28 20:45:09 | [diff] [blame] | 35 | #include "services/screen_ai/screen_ai_library_wrapper_fake.h" |
Ramin Halavati | 767b8fc | 2024-02-02 06:23:08 | [diff] [blame] | 36 | #else |
Abigail Klein | 7a63c57 | 2024-02-28 20:45:09 | [diff] [blame] | 37 | #include "services/screen_ai/screen_ai_library_wrapper_impl.h" |
Ramin Halavati | 767b8fc | 2024-02-02 06:23:08 | [diff] [blame] | 38 | #endif |
| 39 | |
Ramin Halavati | 69bf775 | 2022-04-04 09:58:15 | [diff] [blame] | 40 | namespace screen_ai { |
Ramin Halavati | 26dcce2 | 2022-02-23 13:11:14 | [diff] [blame] | 41 | |
Ramin Halavati | 6d4fc2b | 2022-06-16 15:33:35 | [diff] [blame] | 42 | namespace { |
| 43 | |
Ramin Halavati | 2acf236 | 2024-08-02 14:22:53 | [diff] [blame] | 44 | // How often it would be checked that the service is idle and can be shutdown. |
Ramin Halavati | dc0b2e51 | 2024-08-16 17:39:09 | [diff] [blame] | 45 | constexpr base::TimeDelta kIdleCheckingDelay = base::Minutes(5); |
Ramin Halavati | 2acf236 | 2024-08-02 14:22:53 | [diff] [blame] | 46 | |
Ramin Halavati | 6ff5641 | 2024-08-13 07:15:54 | [diff] [blame] | 47 | // How long after all clients are disconnected, it is checked if service is |
| 48 | // idle. |
Ramin Halavati | 2fa45b8 | 2024-08-15 16:46:53 | [diff] [blame] | 49 | constexpr base::TimeDelta kCoolDownTime = base::Seconds(10); |
Ramin Halavati | 6ff5641 | 2024-08-13 07:15:54 | [diff] [blame] | 50 | |
Ramin Halavati | ad969ef | 2024-05-21 06:59:47 | [diff] [blame] | 51 | // These values are persisted to logs. Entries should not be renumbered and |
| 52 | // numeric values should never be reused. |
Ramin Halavati | 29cc115 | 2024-10-27 17:30:28 | [diff] [blame] | 53 | // See `screen_ai_service.mojom` for more info. |
Ramin Halavati | ad969ef | 2024-05-21 06:59:47 | [diff] [blame] | 54 | enum class OcrClientTypeForMetrics { |
| 55 | kTest = 0, |
| 56 | kPdfViewer = 1, |
| 57 | kLocalSearch = 2, |
| 58 | kCameraApp = 3, |
Ramin Halavati | 29cc115 | 2024-10-27 17:30:28 | [diff] [blame] | 59 | kNotUsed = 4, // Can be used for a new client. |
Ramin Halavati | ad969ef | 2024-05-21 06:59:47 | [diff] [blame] | 60 | kMediaApp = 5, |
Michelle Chen | 5192d52 | 2024-10-22 23:38:22 | [diff] [blame] | 61 | kScreenshotTextDetection, |
| 62 | kMaxValue = kScreenshotTextDetection |
Ramin Halavati | ad969ef | 2024-05-21 06:59:47 | [diff] [blame] | 63 | }; |
| 64 | |
| 65 | OcrClientTypeForMetrics 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 Halavati | ad969ef | 2024-05-21 06:59:47 | [diff] [blame] | 76 | case mojom::OcrClientType::kMediaApp: |
| 77 | return OcrClientTypeForMetrics::kMediaApp; |
Michelle Chen | 5192d52 | 2024-10-22 23:38:22 | [diff] [blame] | 78 | case mojom::OcrClientType::kScreenshotTextDetection: |
| 79 | return OcrClientTypeForMetrics::kScreenshotTextDetection; |
Ramin Halavati | ad969ef | 2024-05-21 06:59:47 | [diff] [blame] | 80 | } |
| 81 | } |
| 82 | |
Ramin Halavati | 89a36c9 | 2023-05-05 15:01:03 | [diff] [blame] | 83 | ui::AXTreeUpdate ConvertVisualAnnotationToTreeUpdate( |
Ramin Halavati | dcb09a17 | 2024-07-15 08:28:55 | [diff] [blame] | 84 | std::optional<chrome_screen_ai::VisualAnnotation>& annotation_proto, |
Ramin Halavati | 89a36c9 | 2023-05-05 15:01:03 | [diff] [blame] | 85 | 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 Klein | 1867370 | 2024-03-05 20:59:06 | [diff] [blame] | 94 | ui::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 Halavati | 26ae6b7 | 2022-11-10 06:45:05 | [diff] [blame] | 103 | } // namespace |
| 104 | |
Ramin Halavati | 39bab0d | 2024-01-30 06:05:01 | [diff] [blame] | 105 | // The library accepts simple pointers to model data retrieval functions, hence |
| 106 | // callback functions with linked object are not safe to pass. |
Ramin Halavati | 63d9c25 | 2024-07-29 06:59:17 | [diff] [blame] | 107 | // This global variable keeps the pointer the only instance of this class. |
| 108 | ModelDataHolder* g_model_data_holder_instance = nullptr; |
Ramin Halavati | 8ced639 | 2023-09-25 07:27:50 | [diff] [blame] | 109 | |
Ramin Halavati | 63d9c25 | 2024-07-29 06:59:17 | [diff] [blame] | 110 | // Keeps the handles of model files, and replies to calls for copying their |
| 111 | // content. |
| 112 | class ModelDataHolder { |
Ramin Halavati | 8ced639 | 2023-09-25 07:27:50 | [diff] [blame] | 113 | public: |
Ramin Halavati | 63d9c25 | 2024-07-29 06:59:17 | [diff] [blame] | 114 | ModelDataHolder() { |
| 115 | CHECK(!g_model_data_holder_instance); |
| 116 | g_model_data_holder_instance = this; |
| 117 | } |
Ramin Halavati | 8ced639 | 2023-09-25 07:27:50 | [diff] [blame] | 118 | |
Ramin Halavati | 63d9c25 | 2024-07-29 06:59:17 | [diff] [blame] | 119 | 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 Halavati | 8ced639 | 2023-09-25 07:27:50 | [diff] [blame] | 125 | } |
| 126 | |
| 127 | // Returns 0 if file is not found. |
| 128 | static uint32_t GetDataSize(const char* relative_file_path) { |
Ramin Halavati | 63d9c25 | 2024-07-29 06:59:17 | [diff] [blame] | 129 | 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 Halavati | 8ced639 | 2023-09-25 07:27:50 | [diff] [blame] | 133 | } |
| 134 | |
Ramin Halavati | 63d9c25 | 2024-07-29 06:59:17 | [diff] [blame] | 135 | // 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 Halavati | 8ced639 | 2023-09-25 07:27:50 | [diff] [blame] | 137 | static void CopyData(const char* relative_file_path, |
| 138 | uint32_t buffer_size, |
| 139 | char* buffer) { |
Ramin Halavati | 63d9c25 | 2024-07-29 06:59:17 | [diff] [blame] | 140 | 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 Sepez | e9b4dd55 | 2024-08-14 22:44:29 | [diff] [blame] | 147 | CHECK_EQ(UNSAFE_TODO(model_file->Read(0, buffer, length)), length); |
Ramin Halavati | 8ced639 | 2023-09-25 07:27:50 | [diff] [blame] | 148 | } |
| 149 | |
Ramin Halavati | 63d9c25 | 2024-07-29 06:59:17 | [diff] [blame] | 150 | 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 Halavati | 39bab0d | 2024-01-30 06:05:01 | [diff] [blame] | 154 | } |
| 155 | } |
Ramin Halavati | 8ced639 | 2023-09-25 07:27:50 | [diff] [blame] | 156 | |
Ramin Halavati | 63d9c25 | 2024-07-29 06:59:17 | [diff] [blame] | 157 | // 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 Halavati | 8ced639 | 2023-09-25 07:27:50 | [diff] [blame] | 165 | private: |
Ramin Halavati | 63d9c25 | 2024-07-29 06:59:17 | [diff] [blame] | 166 | std::map<std::string, base::File> model_files_; |
Ramin Halavati | 8ced639 | 2023-09-25 07:27:50 | [diff] [blame] | 167 | }; |
| 168 | |
Ramin Halavati | 26ae6b7 | 2022-11-10 06:45:05 | [diff] [blame] | 169 | ScreenAIService::ScreenAIService( |
Ramin Halavati | c1e4fa9 | 2023-05-17 17:22:06 | [diff] [blame] | 170 | mojo::PendingReceiver<mojom::ScreenAIServiceFactory> receiver) |
| 171 | : factory_receiver_(this, std::move(receiver)), |
| 172 | ocr_receiver_(this), |
Ramin Halavati | ad969ef | 2024-05-21 06:59:47 | [diff] [blame] | 173 | main_content_extraction_receiver_(this) { |
Ramin Halavati | 6ff5641 | 2024-08-13 07:15:54 | [diff] [blame] | 174 | 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 Halavati | 63d9c25 | 2024-07-29 06:59:17 | [diff] [blame] | 180 | model_data_holder_ = std::make_unique<ModelDataHolder>(); |
Ramin Halavati | 2acf236 | 2024-08-02 14:22:53 | [diff] [blame] | 181 | idle_checking_timer_ = std::make_unique<base::RepeatingTimer>(); |
| 182 | idle_checking_timer_->Start(FROM_HERE, kIdleCheckingDelay, this, |
| 183 | &ScreenAIService::ShutDownIfNoClients); |
Ramin Halavati | ad969ef | 2024-05-21 06:59:47 | [diff] [blame] | 184 | } |
Ramin Halavati | 26ae6b7 | 2022-11-10 06:45:05 | [diff] [blame] | 185 | |
| 186 | ScreenAIService::~ScreenAIService() = default; |
| 187 | |
Ramin Halavati | 8ced639 | 2023-09-25 07:27:50 | [diff] [blame] | 188 | void ScreenAIService::LoadLibrary(const base::FilePath& library_path) { |
Abigail Klein | 7a63c57 | 2024-02-28 20:45:09 | [diff] [blame] | 189 | // The ScopedBlockingCall in LoadLibrary guarantees that this is not run on |
| 190 | // the UI thread. |
Ramin Halavati | a9b50a10 | 2024-02-07 18:37:23 | [diff] [blame] | 191 | #if BUILDFLAG(USE_FAKE_SCREEN_AI) |
Ramin Halavati | 767b8fc | 2024-02-02 06:23:08 | [diff] [blame] | 192 | library_ = std::make_unique<ScreenAILibraryWrapperFake>(); |
| 193 | #else |
| 194 | library_ = std::make_unique<ScreenAILibraryWrapperImpl>(); |
| 195 | #endif |
Ramin Halavati | 8ced639 | 2023-09-25 07:27:50 | [diff] [blame] | 196 | |
| 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 Halavati | 63d9c25 | 2024-07-29 06:59:17 | [diff] [blame] | 220 | library_->SetFileContentFunctions(&ModelDataHolder::GetDataSize, |
| 221 | &ModelDataHolder::CopyData); |
Ramin Halavati | 8ced639 | 2023-09-25 07:27:50 | [diff] [blame] | 222 | } |
| 223 | |
Ramin Halavati | c1e4fa9 | 2023-05-17 17:22:06 | [diff] [blame] | 224 | void ScreenAIService::InitializeMainContentExtraction( |
Ramin Halavati | d139a0f | 2023-02-28 13:25:28 | [diff] [blame] | 225 | const base::FilePath& library_path, |
Ramin Halavati | fbc6202 | 2023-11-10 11:05:07 | [diff] [blame] | 226 | base::flat_map<base::FilePath, base::File> model_files, |
Ramin Halavati | c1e4fa9 | 2023-05-17 17:22:06 | [diff] [blame] | 227 | mojo::PendingReceiver<mojom::MainContentExtractionService> |
| 228 | main_content_extractor_service_receiver, |
| 229 | InitializeMainContentExtractionCallback callback) { |
| 230 | if (!library_) { |
Ramin Halavati | 8ced639 | 2023-09-25 07:27:50 | [diff] [blame] | 231 | LoadLibrary(library_path); |
Ramin Halavati | c1e4fa9 | 2023-05-17 17:22:06 | [diff] [blame] | 232 | } |
| 233 | |
| 234 | if (!library_) { |
| 235 | std::move(callback).Run(false); |
| 236 | base::Process::TerminateCurrentProcessImmediately(-1); |
| 237 | } |
| 238 | |
Ramin Halavati | 63d9c25 | 2024-07-29 06:59:17 | [diff] [blame] | 239 | model_data_holder_->AddModelFiles(std::move(model_files)); |
Ramin Halavati | b044504d | 2022-10-24 06:36:34 | [diff] [blame] | 240 | |
Ramin Halavati | 8ced639 | 2023-09-25 07:27:50 | [diff] [blame] | 241 | bool init_successful = library_->InitMainContentExtraction(); |
Ramin Halavati | 1d57c2d | 2023-05-24 05:18:24 | [diff] [blame] | 242 | base::UmaHistogramBoolean( |
| 243 | "Accessibility.ScreenAI.MainContentExtraction.Initialized", |
| 244 | init_successful); |
| 245 | if (!init_successful) { |
Ramin Halavati | c1e4fa9 | 2023-05-17 17:22:06 | [diff] [blame] | 246 | std::move(callback).Run(false); |
Ramin Halavati | 89a36c9 | 2023-05-05 15:01:03 | [diff] [blame] | 247 | return; |
Ramin Halavati | d139a0f | 2023-02-28 13:25:28 | [diff] [blame] | 248 | } |
Ramin Halavati | 89a36c9 | 2023-05-05 15:01:03 | [diff] [blame] | 249 | |
Ramin Halavati | c1e4fa9 | 2023-05-17 17:22:06 | [diff] [blame] | 250 | // 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 Halavati | 2acf236 | 2024-08-02 14:22:53 | [diff] [blame] | 257 | main_content_extraction_last_used_ = base::TimeTicks::Now(); |
Ramin Halavati | c1e4fa9 | 2023-05-17 17:22:06 | [diff] [blame] | 258 | } |
| 259 | |
| 260 | void ScreenAIService::InitializeOCR( |
| 261 | const base::FilePath& library_path, |
Ramin Halavati | fbc6202 | 2023-11-10 11:05:07 | [diff] [blame] | 262 | base::flat_map<base::FilePath, base::File> model_files, |
Ramin Halavati | c1e4fa9 | 2023-05-17 17:22:06 | [diff] [blame] | 263 | mojo::PendingReceiver<mojom::OCRService> ocr_service_receiver, |
| 264 | InitializeOCRCallback callback) { |
| 265 | if (!library_) { |
Ramin Halavati | 8ced639 | 2023-09-25 07:27:50 | [diff] [blame] | 266 | LoadLibrary(library_path); |
Ramin Halavati | c1e4fa9 | 2023-05-17 17:22:06 | [diff] [blame] | 267 | } |
| 268 | |
| 269 | if (!library_) { |
| 270 | std::move(callback).Run(false); |
| 271 | base::Process::TerminateCurrentProcessImmediately(-1); |
| 272 | } |
| 273 | |
Ramin Halavati | 63d9c25 | 2024-07-29 06:59:17 | [diff] [blame] | 274 | model_data_holder_->AddModelFiles(std::move(model_files)); |
| 275 | |
Ramin Halavati | 0e2ae574 | 2023-11-10 05:57:54 | [diff] [blame] | 276 | bool init_successful = library_->InitOCR(); |
Ramin Halavati | 35d129f | 2023-06-22 16:37:09 | [diff] [blame] | 277 | base::UmaHistogramBoolean("Accessibility.ScreenAI.OCR.Initialized", |
Ramin Halavati | 1d57c2d | 2023-05-24 05:18:24 | [diff] [blame] | 278 | init_successful); |
Ramin Halavati | 48058433 | 2023-08-02 06:38:40 | [diff] [blame] | 279 | |
Ramin Halavati | 1d57c2d | 2023-05-24 05:18:24 | [diff] [blame] | 280 | if (!init_successful) { |
Ramin Halavati | c1e4fa9 | 2023-05-17 17:22:06 | [diff] [blame] | 281 | 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 Halavati | 2acf236 | 2024-08-02 14:22:53 | [diff] [blame] | 291 | ocr_last_used_ = base::TimeTicks::Now(); |
Ramin Halavati | b044504d | 2022-10-24 06:36:34 | [diff] [blame] | 292 | } |
Ramin Halavati | 26dcce2 | 2022-02-23 13:11:14 | [diff] [blame] | 293 | |
| 294 | void ScreenAIService::BindAnnotator( |
| 295 | mojo::PendingReceiver<mojom::ScreenAIAnnotator> annotator) { |
| 296 | screen_ai_annotators_.Add(this, std::move(annotator)); |
| 297 | } |
| 298 | |
Ramin Halavati | eddadb6 | 2022-05-04 17:29:49 | [diff] [blame] | 299 | void ScreenAIService::BindMainContentExtractor( |
| 300 | mojo::PendingReceiver<mojom::Screen2xMainContentExtractor> |
| 301 | main_content_extractor) { |
Ramin Halavati | 6ff5641 | 2024-08-13 07:15:54 | [diff] [blame] | 302 | screen2x_main_content_extractors_.Add(this, |
| 303 | std::move(main_content_extractor)); |
Ramin Halavati | eddadb6 | 2022-05-04 17:29:49 | [diff] [blame] | 304 | } |
| 305 | |
Arthur Sonzogni | c571efb | 2024-01-26 20:26:18 | [diff] [blame] | 306 | std::optional<chrome_screen_ai::VisualAnnotation> |
Ramin Halavati | 7716b03 | 2024-08-19 14:56:20 | [diff] [blame] | 307 | ScreenAIService::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 Halavati | ad969ef | 2024-05-21 06:59:47 | [diff] [blame] | 312 | base::UmaHistogramEnumeration("Accessibility.ScreenAI.OCR.ClientType", |
Ramin Halavati | 7716b03 | 2024-08-19 14:56:20 | [diff] [blame] | 313 | GetClientType(client_type)); |
Ramin Halavati | ad969ef | 2024-05-21 06:59:47 | [diff] [blame] | 314 | |
Ramin Halavati | 2acf236 | 2024-08-02 14:22:53 | [diff] [blame] | 315 | ocr_last_used_ = base::TimeTicks::Now(); |
Ramin Halavati | 7252cd0 | 2023-05-10 07:21:33 | [diff] [blame] | 316 | auto result = library_->PerformOcr(image); |
Ramin Halavati | 2acf236 | 2024-08-02 14:22:53 | [diff] [blame] | 317 | base::TimeDelta elapsed_time = base::TimeTicks::Now() - ocr_last_used_; |
Ramin Halavati | 2650ebe | 2023-11-27 19:56:35 | [diff] [blame] | 318 | int lines_count = result ? result->lines_size() : 0; |
Ramin Halavati | 3ae33ca | 2024-01-01 12:58:39 | [diff] [blame] | 319 | unsigned image_size = image.width() * image.height(); |
Ramin Halavati | 2650ebe | 2023-11-27 19:56:35 | [diff] [blame] | 320 | VLOG(1) << "OCR returned " << lines_count << " lines in " << elapsed_time; |
Ramin Halavati | 7252cd0 | 2023-05-10 07:21:33 | [diff] [blame] | 321 | |
Ramin Halavati | 18aba87 | 2024-08-23 14:33:05 | [diff] [blame] | 322 | if (!result) { |
| 323 | base::UmaHistogramEnumeration( |
| 324 | "Accessibility.ScreenAI.OCR.Failed.ClientType", |
| 325 | GetClientType(client_type)); |
| 326 | } |
Ramin Halavati | 81b6d92 | 2024-06-07 14:20:34 | [diff] [blame] | 327 | base::UmaHistogramBoolean("Accessibility.ScreenAI.OCR.Successful", |
| 328 | result.has_value()); |
Ramin Halavati | 2650ebe | 2023-11-27 19:56:35 | [diff] [blame] | 329 | base::UmaHistogramCounts100("Accessibility.ScreenAI.OCR.LinesCount", |
| 330 | lines_count); |
Ramin Halavati | a0548b26 | 2023-06-27 05:56:47 | [diff] [blame] | 331 | base::UmaHistogramCounts10M("Accessibility.ScreenAI.OCR.ImageSize10M", |
Ramin Halavati | 3ae33ca | 2024-01-01 12:58:39 | [diff] [blame] | 332 | 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 Halavati | 2650ebe | 2023-11-27 19:56:35 | [diff] [blame] | 346 | |
Ramin Halavati | 7716b03 | 2024-08-19 14:56:20 | [diff] [blame] | 347 | // MediaApp provides OCR for ChromeOS PDF viewer. |
| 348 | if (client_type == mojom::OcrClientType::kPdfViewer || |
| 349 | client_type == mojom::OcrClientType::kMediaApp) { |
Ramin Halavati | 2650ebe | 2023-11-27 19:56:35 | [diff] [blame] | 350 | 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 Halavati | 4c47eb8 | 2024-12-03 19:06:42 | [diff] [blame^] | 356 | |
| 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 Halavati | 2650ebe | 2023-11-27 19:56:35 | [diff] [blame] | 364 | } |
| 365 | |
Ramin Halavati | 7252cd0 | 2023-05-10 07:21:33 | [diff] [blame] | 366 | return result; |
| 367 | } |
| 368 | |
Ramin Halavati | ad969ef | 2024-05-21 06:59:47 | [diff] [blame] | 369 | void ScreenAIService::SetClientType(mojom::OcrClientType client_type) { |
| 370 | ocr_client_types_[screen_ai_annotators_.current_receiver()] = client_type; |
| 371 | } |
| 372 | |
Dmitry Grebenyuk | c49b95ee | 2023-05-02 05:52:07 | [diff] [blame] | 373 | void ScreenAIService::PerformOcrAndReturnAnnotation( |
| 374 | const SkBitmap& image, |
| 375 | PerformOcrAndReturnAnnotationCallback callback) { |
Arthur Sonzogni | c571efb | 2024-01-26 20:26:18 | [diff] [blame] | 376 | std::optional<chrome_screen_ai::VisualAnnotation> annotation_proto = |
Ramin Halavati | 7716b03 | 2024-08-19 14:56:20 | [diff] [blame] | 377 | PerformOcrAndRecordMetrics(image); |
Dmitry Grebenyuk | c49b95ee | 2023-05-02 05:52:07 | [diff] [blame] | 378 | |
Ramin Halavati | 89a36c9 | 2023-05-05 15:01:03 | [diff] [blame] | 379 | if (annotation_proto) { |
| 380 | std::move(callback).Run(ConvertProtoToVisualAnnotation(*annotation_proto)); |
| 381 | return; |
Dmitry Grebenyuk | c49b95ee | 2023-05-02 05:52:07 | [diff] [blame] | 382 | } |
Ramin Halavati | 89a36c9 | 2023-05-05 15:01:03 | [diff] [blame] | 383 | |
| 384 | std::move(callback).Run(mojom::VisualAnnotation::New()); |
Dmitry Grebenyuk | c49b95ee | 2023-05-02 05:52:07 | [diff] [blame] | 385 | } |
| 386 | |
Kyungjun Lee | 9046152 | 2023-04-25 06:50:09 | [diff] [blame] | 387 | void ScreenAIService::PerformOcrAndReturnAXTreeUpdate( |
Ramin Halavati | b446c02 | 2023-03-29 15:36:36 | [diff] [blame] | 388 | const SkBitmap& image, |
Kyungjun Lee | 9046152 | 2023-04-25 06:50:09 | [diff] [blame] | 389 | PerformOcrAndReturnAXTreeUpdateCallback callback) { |
Arthur Sonzogni | c571efb | 2024-01-26 20:26:18 | [diff] [blame] | 390 | std::optional<chrome_screen_ai::VisualAnnotation> annotation_proto = |
Ramin Halavati | 7716b03 | 2024-08-19 14:56:20 | [diff] [blame] | 391 | PerformOcrAndRecordMetrics(image); |
Ramin Halavati | 89a36c9 | 2023-05-05 15:01:03 | [diff] [blame] | 392 | ui::AXTreeUpdate update = ConvertVisualAnnotationToTreeUpdate( |
| 393 | annotation_proto, gfx::Rect(image.width(), image.height())); |
Kyungjun Lee | 9046152 | 2023-04-25 06:50:09 | [diff] [blame] | 394 | |
Ramin Halavati | 89a36c9 | 2023-05-05 15:01:03 | [diff] [blame] | 395 | // 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 Halavati | b446c02 | 2023-03-29 15:36:36 | [diff] [blame] | 398 | } |
| 399 | |
| 400 | void ScreenAIService::ExtractMainContent(const ui::AXTreeUpdate& snapshot, |
Ramin Halavati | b446c02 | 2023-03-29 15:36:36 | [diff] [blame] | 401 | ExtractMainContentCallback callback) { |
Ramin Halavati | 2acf236 | 2024-08-02 14:22:53 | [diff] [blame] | 402 | main_content_extraction_last_used_ = base::TimeTicks::Now(); |
Abigail Klein | 1867370 | 2024-03-05 20:59:06 | [diff] [blame] | 403 | ui::AXTree tree; |
| 404 | std::optional<std::vector<int32_t>> content_node_ids; |
| 405 | bool success = ExtractMainContentInternal(snapshot, tree, content_node_ids); |
Abigail Klein | 1867370 | 2024-03-05 20:59:06 | [diff] [blame] | 406 | |
| 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 | |
| 414 | void 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 | |
| 428 | bool ScreenAIService::ExtractMainContentInternal( |
| 429 | const ui::AXTreeUpdate& snapshot, |
| 430 | ui::AXTree& tree, |
| 431 | std::optional<std::vector<int32_t>>& content_node_ids) { |
Ramin Halavati | b446c02 | 2023-03-29 15:36:36 | [diff] [blame] | 432 | // Early return if input is empty. |
| 433 | if (snapshot.nodes.empty()) { |
Abigail Klein | 1867370 | 2024-03-05 20:59:06 | [diff] [blame] | 434 | return false; |
Ramin Halavati | b33fc0c | 2022-05-06 09:32:22 | [diff] [blame] | 435 | } |
Ramin Halavati | e2549a9 | 2022-08-02 07:43:19 | [diff] [blame] | 436 | |
Abigail Klein | 1867370 | 2024-03-05 20:59:06 | [diff] [blame] | 437 | // Deserialize the snapshot and reserialize it to a view hierarchy proto. |
| 438 | CHECK(tree.Unserialize(snapshot)); |
Ramin Halavati | 62e034d | 2024-08-02 08:58:22 | [diff] [blame] | 439 | std::optional<ViewHierarchyAndTreeSize> converted_snapshot = |
| 440 | SnapshotToViewHierarchy(tree); |
| 441 | if (!converted_snapshot) { |
| 442 | VLOG(0) << "Proto not generated."; |
| 443 | return false; |
| 444 | } |
| 445 | |
Ramin Halavati | 62e034d | 2024-08-02 08:58:22 | [diff] [blame] | 446 | content_node_ids = |
| 447 | library_->ExtractMainContent(converted_snapshot->serialized_proto); |
Ramin Halavati | 81b6d92 | 2024-06-07 14:20:34 | [diff] [blame] | 448 | base::UmaHistogramBoolean( |
| 449 | "Accessibility.ScreenAI.MainContentExtraction.Successful", |
| 450 | content_node_ids.has_value()); |
Abigail Klein | 1867370 | 2024-03-05 20:59:06 | [diff] [blame] | 451 | if (content_node_ids.has_value() && content_node_ids->size() > 0) { |
Ramin Halavati | 5aad701d | 2023-05-09 17:23:29 | [diff] [blame] | 452 | VLOG(2) << "Screen2x returned " << content_node_ids->size() << " node ids."; |
Abigail Klein | 1867370 | 2024-03-05 20:59:06 | [diff] [blame] | 453 | return true; |
| 454 | } else { |
| 455 | VLOG(0) << "Screen2x returned no results."; |
| 456 | return false; |
Ramin Halavati | 5aad701d | 2023-05-09 17:23:29 | [diff] [blame] | 457 | } |
Abigail Klein | 1867370 | 2024-03-05 20:59:06 | [diff] [blame] | 458 | } |
Ramin Halavati | 89a36c9 | 2023-05-05 15:01:03 | [diff] [blame] | 459 | |
Abigail Klein | 1867370 | 2024-03-05 20:59:06 | [diff] [blame] | 460 | ui::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 Klein | dfdde35 | 2023-01-27 21:03:10 | [diff] [blame] | 464 | } |
| 465 | |
Ramin Halavati | 6ff5641 | 2024-08-13 07:15:54 | [diff] [blame] | 466 | void ScreenAIService::OcrReceiverDisconnected() { |
Ramin Halavati | ad969ef | 2024-05-21 06:59:47 | [diff] [blame] | 467 | 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 Halavati | 6ff5641 | 2024-08-13 07:15:54 | [diff] [blame] | 471 | |
| 472 | CheckIdleStateAfterDelay(); |
| 473 | } |
| 474 | |
| 475 | void 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 Halavati | ad969ef | 2024-05-21 06:59:47 | [diff] [blame] | 482 | } |
| 483 | |
Ramin Halavati | 3842e7e | 2024-07-31 04:39:21 | [diff] [blame] | 484 | void ScreenAIService::ShutDownIfNoClients() { |
Ramin Halavati | 2acf236 | 2024-08-02 14:22:53 | [diff] [blame] | 485 | const base::TimeTicks kIdlenessThreshold = |
| 486 | base::TimeTicks::Now() - kIdleCheckingDelay; |
Ramin Halavati | dc0b2e51 | 2024-08-16 17:39:09 | [diff] [blame] | 487 | 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 Halavati | 2acf236 | 2024-08-02 14:22:53 | [diff] [blame] | 496 | } |
Ramin Halavati | 3842e7e | 2024-07-31 04:39:21 | [diff] [blame] | 497 | } |
| 498 | |
Ramin Halavati | 26dcce2 | 2022-02-23 13:11:14 | [diff] [blame] | 499 | } // namespace screen_ai |