Installing TensorFlow for CUDA 8

Andrew Silva | 23 April 2018

TensorFlow is a fast and useful deep learning library, but the documentation is a bit lacking for older versions of CUDA or CuDNN. This is complicated if you want to use something like Keras, which now depends rather heavily on TensorFlow. Fortunately, the process is just as easy, and you can get off the ground in no time. If you’re on CUDA 8, you simply need to refer to the older binary of TensorFlow, specifically TF 1.4. You can find the install documentation here, but the short version is to get into whichever environment you are planning on using (virtualenvs...

read more

Face Detection with Caffe in Python

Andrew Silva | 20 January 2017

I’ve been working on face detection lately, and Caffe has been tremendously helpful (see my last post for how that installation went…) and provided me with the ability to build on the incredible work available here: https://github.com/kpzhang93/MTCNN_face_detection_alignment

read more

Installing Caffe on macOS

Andrew Silva | 19 January 2017

I recently had to install Caffe on a new MacBook Pro, and it was something of an ordeal. I decided to put a short guide up here in the hope that it will make life easier for anyone else trying to get up and running with Caffe but being stuck with some outdated docs.

read more

Speech Recognition with Python on a Mac

Andrew Silva | 13 November 2016

Setting up speech recognition and transcription on a Mac is surprisingly easy using Python. I’m using Python 3.5, but the library I’ll be using works for 2.7 and 3.x.

read more

Recognizing Facial Expressions with OpenCV 3+

Andrew Silva | 09 November 2016

Facial expressions are a very interesting problem for a variety of reasons. Of interest to me particularly is trying to mirror facial expressions in robotics, to make people more comfortable or to ease communication between humans and robots. As a starting point, if you want to train your own classifiers on the small dataset that my project references, you’ll need to work with the Cohn and Kanade dataset, available here. Similarly, the “giant dataset” that my code refers to is the fer2013.csv file that you can download from Kaggle. However, you can just use the models bundled up in my...

read more

Raspberry Pi Magic Mirror

Andrew Silva | 09 November 2016

Magic Mirrors have been all the rage lately, and I couldn’t resist the urge to make one myself after seeing so many different iterations come out. I had an old Raspberry Pi laying around and a couple of unused small TVs that I could repurpose, so I figured I was already halfway to success.

read more