{"id":20619727,"url":"https://github.com/twtrubiks/face-recognition-notes","last_synced_at":"2025-04-15T12:02:08.120Z","repository":{"id":84518947,"uuid":"79425009","full_name":"twtrubiks/face-recognition-notes","owner":"twtrubiks","description":"紀錄 face-recognition 學習","archived":false,"fork":false,"pushed_at":"2017-09-24T07:02:31.000Z","size":2,"stargazers_count":6,"open_issues_count":0,"forks_count":2,"subscribers_count":2,"default_branch":"master","last_synced_at":"2025-03-28T19:53:43.992Z","etag":null,"topics":["bosssensor","face-recognition","notes","opencv","tensorflow"],"latest_commit_sha":null,"homepage":"","language":null,"has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/twtrubiks.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2017-01-19T06:56:54.000Z","updated_at":"2021-07-05T06:40:21.000Z","dependencies_parsed_at":"2023-03-02T04:30:28.214Z","dependency_job_id":null,"html_url":"https://github.com/twtrubiks/face-recognition-notes","commit_stats":null,"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/twtrubiks%2Fface-recognition-notes","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/twtrubiks%2Fface-recognition-notes/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/twtrubiks%2Fface-recognition-notes/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/twtrubiks%2Fface-recognition-notes/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/twtrubiks","download_url":"https://codeload.github.com/twtrubiks/face-recognition-notes/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":249067779,"owners_count":21207395,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["bosssensor","face-recognition","notes","opencv","tensorflow"],"created_at":"2024-11-16T12:12:22.797Z","updated_at":"2025-04-15T12:02:08.114Z","avatar_url":"https://github.com/twtrubiks.png","language":null,"funding_links":[],"categories":[],"sub_categories":[],"readme":"# face-recognition-notes\n\n## 人臉辨識 ( face recognition ) - 使用 深度學習 ( TensorFlow ) 和 OpenCV\n\n這些是我之前有興趣做的簡單筆記，大家可以參考看看 :smiley:\n\n### 深度學習\n\n一個函數集，自己訓練出來的(經過大量的訓練過程，最終計算出一個最佳函數，得出最佳解，我們主要提供他規則)\n\n### 為什麼要研究深度學習 ? 單純使用 opencv 沒辦到辦到 人臉辨識嗎 ?\n\n可以 。可以透過 opencv 的 Face Recognition 辦到臉部辨識。\n\nopencv 的 Face Recognition 可參考 [FaceDetect](https://github.com/twtrubiks/FaceDetect#後記)\n\n### 那為什麼在網路上找 Face Recognition ，相關近期資料卻都是 深度學習(TensorFlow) 搭配 OpenCV ?\n\n目前猜測是辨識精準度的問題。\n\n單純使用 OpenCV 完成臉部辨識 : 使用人工的方式截取臉部特徵值\n\n使用 OpenCV +  深度學習(TensorFlow) 完成臉部辨識 : 深度學習會自動截取臉部特徵值\n\n考慮到辨識對象可能長的很相似，所以對辨識精準度的要求要比較高。\n\n\n因為這些原因，所以我研究臉部辨識這議題，我是針對深度學習下去研究，opencv 單純只是幫我偵測出人臉。\n\n### 研究專案\n\n參考專案 : [BossSensor](http://ahogrammer.com/2016/11/15/deep-learning-enables-you-to-hide-screen-when-your-boss-is-approaching/)\n\nBossSensor 流程 :\n\n**步驟一** : 收集圖片，圖片只保留臉部的部份，因為如果有背景以及其他不必要的部份，會影響訓練以及辨識.\n\n方法: 使用opencv截取人臉的部份。(在opencv裡，可以透過 haarcascade_frontalface_default.xml 來偵測出目前是否為人臉)\n\n**步驟二** : 開始建立 機器學習(Machine Learning) 的 model\n\n方法 : 透過 Keras  卷積深度神經網路（Convolutional Neuron Networks, CNN） 建立 model\n\n大家為什麼選 卷積深度神經網路（Convolutional Neuron Networks, CNN） ?\n\n因為卷積神經網路 (深度學習結構)在 **圖像** 和 **語音辨識** 方面能夠給出更優的結果。\n\n* Keras 介紹\n\nKeras的後端有 TensorFlow(tf) 以及 Theano(th)\n\nTensorFlow(tf) 表達模式 - (100,3,16,32)\n\nTheano(th) 表達模式 - (100,16,32,3)\n\n100,3 這部份是指 100張 RPG 三通道的圖片\n\n16代表高   32代表寬\n\n訓練和測試使用同一種後端。\n\n在訓練的時候，我們會先把照片轉成大小一致(64*64)\n\ndataset --\u003e build Model --\u003e train Model  --\u003e save Model\n\n**步驟三** : 開始辨識\n\nload Model --\u003e 透過相機截取臉部 --\u003e 將大小轉成(64*64) --\u003e 比對 (開始辨識)\n\n### 相關議題  Microsoft Face API 背後技術\n\n由於 Microsoft Face API 的辨識速度很快，而且上傳給他的圖片也只有幾張而已，猜測他們的技術是比對兩張圖片的特徵值，\n\n而不是使用深度學習\n\n### 目前深度學習的應用\n\n以 google舉例\n\n* Gmail 自動判斷約99%的垃圾郵件\n\n* Google Now 的語音辨識，透過學習，辨識率越來越精準\n\n* Google 相簿的自動分類 (對照片自動下標籤)\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Ftwtrubiks%2Fface-recognition-notes","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Ftwtrubiks%2Fface-recognition-notes","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Ftwtrubiks%2Fface-recognition-notes/lists"}