What is this?
OpenFace is a free, open-source face recognition toolkit built on deep neural networks, developed by Carnegie Mellon University.
OpenFace is a free, open-source face recognition toolkit built on deep neural networks, developed by Carnegie Mellon University.Is this a ready-made app or something I plug into my own code?It is mainly a Python library (the 'openface' package) backed by a Torch/Lua neural network, plus standalone demo scripts for comparison, classification and a real-time web demo rather than one polished application.
It is mainly a Python library (the 'openface' package) backed by a Torch/Lua neural network, plus standalone demo scripts for comparison, classification and a real-time web demo rather than one polished application.How do I try it out?Start with the scripts in demos/: 'compare.py' checks if two photos match, 'classifier.py' trains and runs a classifier, 'classifierwebcam.py' applies a trained classifier to a live webcam feed, and 'batch-represent' generates embeddings for a batch of images.
Start with the scripts in demos/: 'compare.py' checks if two photos match, 'classifier.py' trains and runs a classifier, 'classifierwebcam.py' applies a trained classifier to a live webcam feed, and 'batch-represent' generates embeddings for a batch of images.Do I need Torch and dlib set up before using it?Yes - the recognition network itself runs on Torch (Lua), face alignment reuses dlib's 68-point landmark detector, and a Docker image is provided so you do not have to assemble this stack by hand.
Yes - the recognition network itself runs on Torch (Lua), face alignment reuses dlib's 68-point landmark detector, and a Docker image is provided so you do not have to assemble this stack by hand.Who is this meant for?It's built for researchers and developers who want to experiment with deep-learning face recognition, verification, and clustering, not end users looking for a plug-and-play consumer app.
It's built for researchers and developers who want to experiment with deep-learning face recognition, verification, and clustering, not end users looking for a plug-and-play consumer app.Can I train my own face recognition model instead of using the bundled one?Yes, the training/ directory contains scripts to train new OpenFace neural network models from scratch.
Yes, the training/ directory contains scripts to train new OpenFace neural network models from scratch.How is accuracy measured?The evaluation/ folder has scripts that benchmark the model on the LFW (Labeled Faces in the Wild) dataset.
The evaluation/ folder has scripts that benchmark the model on the LFW (Labeled Faces in the Wild) dataset.Who funded this research?Development was backed by an NSF grant, with additional support from Intel, Google, Vodafone, and NVIDIA.
Development was backed by an NSF grant, with additional support from Intel, Google, Vodafone, and NVIDIA.Do I need to cite OpenFace if I use it in my work?Yes, the README asks that publications using OpenFace cite the project's tech report, and it provides a ready-made BibTeX entry.
Yes, the README asks that publications using OpenFace cite the project's tech report, and it provides a ready-made BibTeX entry.What are similar projects?facenet, deepface, caffe, vit-pytorch, CLIP, DeepLearning, deep-learning-for-image-processing, pix2code
facenet, deepface, caffe, vit-pytorch, CLIP, DeepLearning, deep-learning-for-image-processing, pix2codeWhich topics is it in?face-recognition, deep-neural-networks, classification, image-classification, pretrained-models, lua, academic, evaluation, neural-network, real-time, artificial-intelligence, bsd
face-recognition, deep-neural-networks, classification, image-classification, pretrained-models, lua, academic, evaluation, neural-network, real-time, artificial-intelligence, bsd