datahacker.rs@gmail.com

Category: Deep Learning

#OD1 YOLO Object Detection

YOU ONLY LOOK ONCE Highlights: In this post we will learn about the YOLO Object Detection system, and how to implement such a system with Keras. About Yolo: Our unified architecture is extremely fast. Our base YOLO model processes imagesin real-time at 45 frames per second. A smaller version of the network, Fast YOLO,processes an astounding 155 frames per second … — You Only Look Once: Unified, Real-Time Object Detection, 2015 Tutorial Overview: This post…
Read more

#FA 003 Face Detection with OpenCV in Videos

In real case scenarios, there is often a need for detection and recognition of faces not just in images, but in videos. For example, this is a necessary prerequisite for security cameras, filters on social networks, identification at work and many other cases where cameras are used. The pictures are static, but videos can be seen as a series of pictures, or frames. All video clips are constructed of a constant number of frames in…
Read more

# K An implementation of a Convolutional Neural Network in Keras – MNIST dataset

In this post we will see how we can classify handwritten digits using convolutional neural network implemented in Keras. Required packages: Numpy Matplotlib Keras Tensorflow Sklearn Seaborn __future__ Keras-vis Table of Contents: Load the digit dataset Implementing a Neural Network Visualization Test our model Save a model as picture Activation Maps Saliency maps Activations Now we can plot some predictions, to see how our model works. Images for testing can be downloaded from the internet…
Read more

#025 CNN Bounding Box Predictions

Bounding box predictions In the last post, we learned how to use a convolutional implementation of sliding windows. That’s more computationally efficient, but it still has a problem of not outputting the most accurate bounding boxes.  In this post, we will see how we can obtain more accurate predictions of bounding boxes.  Output accurate bounding boxes With sliding windows, we take the sets of windows that we move throughout the image and we obtain a set of sliding…
Read more

#017 CNN Inception Network

\(Inception\enspace network \)  Motivation for the \(Inception\enspace network \): In the last post we talked about why \(1\times 1 \) convolutional layer can be useful and now we will use it for building the \(Inception\enspace network \). When designing a layer for a \(convnet \) we might have to choose between a \(1 \times 3 \) filter, a \(3 \times 3\) or \(5\times 5\) or maybe a pooling layer. An \(Inception\enspace network \) solves this…
Read more