Difference between revisions of "Openface"
Onnowpurbo (talk | contribs) |
Onnowpurbo (talk | contribs) |
||
| Line 23: | Line 23: | ||
| − | Overview | + | ==Overview Workflow== |
| − | + | Berikut ini adalah overview workflow pendeteksian wajah seseorang yang sudah ada di contoh opencv | |
| − | * | + | * Deteksi wajah dengan model pre-trained dari dlib atau OpenCV. |
| − | * | + | * Transformasikan wajah untuk neural network. OpenFace menggunakan dlib real-time pose estimation dengan OpenCV affine transformation untuk membuat semua mata dan bibir berada pada lokasi yang sama untuk semua gambar. |
* Use a deep neural network to represent (or embed) the face on a 128-dimensional unit hypersphere. The embedding is a generic representation for anybody's face. Unlike other face representations, this embedding has the nice property that a larger distance between two face embeddings means that the faces are likely not of the same person. This property makes clustering, similarity detection, and classification tasks easier than other face recognition techniques where the Euclidean distance between features is not meaningful. | * Use a deep neural network to represent (or embed) the face on a 128-dimensional unit hypersphere. The embedding is a generic representation for anybody's face. Unlike other face representations, this embedding has the nice property that a larger distance between two face embeddings means that the faces are likely not of the same person. This property makes clustering, similarity detection, and classification tasks easier than other face recognition techniques where the Euclidean distance between features is not meaningful. | ||
* Apply your favorite clustering or classification techniques to the features to complete your recognition task. See below for our examples for classification and similarity detection, including an online web demo. | * Apply your favorite clustering or classification techniques to the features to complete your recognition task. See below for our examples for classification and similarity detection, including an online web demo. | ||
| − | |||
==Referensi== | ==Referensi== | ||
Revision as of 06:13, 25 May 2018
Sumber:
- https://github.com/TadasBaltrusaitis/OpenFace/wiki
- https://cmusatyalab.github.io/openface/
- https://blog.algorithmia.com/understanding-facial-recognition-openface/
OpenFace adalah implementasi pengenalan wajah menggunakan Python dan Torch dengan deep neural network dan berdasarkan pada CVPR 2015 paper FaceNet: A Unified Embedding for Face Recognition and Clustering oleh Florian Schroff, Dmitry Kalenichenko, dan James Philbin di Google. Torch memungkinkan jaringan dieksekusi pada CPU atau dengan CUDA.
Dibuat oleh Brandon Amos, Bartosz Ludwiczuk, dan Mahadev Satyanarayanan.
- Open Source, denan source code tersedia di GitHub pada cmusatyalab/openface.
- Ada dokumentasi API
- Bergabung pada group cmu-openface atau gitter chat untuk diskusi dan masalah instanasi.
Penelitian ini di dukung oleh National Science Foundation (NSF) dengan dukungan tambahan dari Intel, Google, Vodafone, NVIDIA, dan Conklin Kistler family fund.
Apakah pengenalan wajah menyelesaikan sebuah masalah?
Tidak! Akurasi dari berbagai makalah penelitian tampaknya baru saja mulai melampaui akurasi manusia pada beberapa standard. Keakuratan sistem pengenalan wajah open source memang masih tertinggal di belakang state-of-the-art.
Gunakan Dengan Bertanggung Jawab
OpenFace tidak mendukung penggunaan aplikasi ini untuk melanggar privacy dan keamanan. Openface digunakan untuk membantuk mereka yang kesulitan dalam pengenalan untuk mengenali dunia sekeliling mereka.
Overview Workflow
Berikut ini adalah overview workflow pendeteksian wajah seseorang yang sudah ada di contoh opencv
- Deteksi wajah dengan model pre-trained dari dlib atau OpenCV.
- Transformasikan wajah untuk neural network. OpenFace menggunakan dlib real-time pose estimation dengan OpenCV affine transformation untuk membuat semua mata dan bibir berada pada lokasi yang sama untuk semua gambar.
- Use a deep neural network to represent (or embed) the face on a 128-dimensional unit hypersphere. The embedding is a generic representation for anybody's face. Unlike other face representations, this embedding has the nice property that a larger distance between two face embeddings means that the faces are likely not of the same person. This property makes clustering, similarity detection, and classification tasks easier than other face recognition techniques where the Euclidean distance between features is not meaningful.
- Apply your favorite clustering or classification techniques to the features to complete your recognition task. See below for our examples for classification and similarity detection, including an online web demo.