SCIENTIFIC VALIDATION
How does it work?
Let's dive in.
SCIENTIFIC VALIDATION
So, how are we able to predict attention with 95% accuracy?
Thanks to the culmination of years of research and experience in the fields of Consumer Neuroscience and AI, we've been able to develop our own unique neural net that can predict visual salience (attention) with a remarkable accuracy of 95% relative to traditional eye tracking.
- 95% Accuracy in Salience
- Data from 10,000+ Images
- ANN-Based Attention Prediction
- Uses GAN and ConvNet Tech
- Optimized Through Training
- AUC Score: 0.87 Validation
- White Paper for Details
03
Defining the network architecture
The attention prediction produced by attention ai is ... read more
01 - It all starts with data
Our training data
The first step in our process is to gather visual saliency data from a large set of visual stimuli, such as advertisement images or video frames. This data is then exposed to a diverse group of real people in an online environment. By simulating eye-tracking, there is no need for this group to participate in a study with real eye-trackers.
Our network was trained on the attention data of thousands of participants. These participants together looked at more than 10,000 images. Each image was seen by at least 50 participants. This resulted in a training data set with a “gold truth” attention heatmap for each image.
Note:Visual saliency refers to the idea that certain parts of a scene are pre-attentively distinctive and create some form of immediate significant visual arousal. In the images the purple dot on the bottom has higher visual salience than the divergent blue dot on the top.
02 - It's ANN for short
About our neural network
The attention predictions on our platform are generated by an Artificial Neural Network (ANN). An ANN is a collection of nodes that are connected. Each connection has a weight determining how much one node impacts the next. The nodes are aggregated into layers.
An ANN is trained using a large set of training data. During training the weights of the network are updated. If trained correctly, it can transform any input into the correct output.
The image visualises the connection of nodes in the neural network between the input and output layers.
03 - Defining the network architecture
What's in between those inputs and outputs?
The attention prediction produced by attention ai is generated using a so-called generative adversarial network, or GAN. The input is the RGB values of your image or video. The output is the attention prediction. On the input side a convolutional neural network, or ConvNet, is used that was pre-trained on over 14 million images to detect objects, and therefore it already holds a latent representation of object identity.
After gathering enough training data, the results are fed to an artificial neural network using a specialized training scheme.
04 - Training the neural net
How do we ensure the neural net knows the 'truth'?
Before training, the GAN has random weights, resulting in faulty heatmaps. The difference between the predicted heatmap and the “gold truth” heatmap is the error. Mathematical procedures subsequently adjust the weights in a way that reduces this error.
After a large number of training rounds, the network converges to producing heatmaps that are more and more similar to the “gold truth”.
Learning how to 'learn'
After gathering enough training data, the results are fed to an artificial neural network, using a specialized training scheme. After optimizing the neural net, it can simulate the saliency results gathered in the first step. Beyond that, the neural net can make accurate saliency predictions of visual material it has never seen before.
