Intelligent Automation for Work and Life
At Quantilus, we have been working with AI before it became cool (and scary). Our first foray was in the field of Natural Language Processing – which we used for automated grammar and style checks of written content. Subsequently, we built tools to classify untagged content in intelligent, usable ways, and to present it for consumption with a high degree of personalization. More recently, we have been working on personality assessment of individuals based on 1) the words they speak (a relatively simple task), and 2) changes in their facial patterns based on verbal and visual cues (a much more complex task).
Want to build Virtual Reality or Augmented Reality apps for your business? We built some of the first business-focused AR apps for mobile and wearable platforms through our SAP partnership. Our apps help technical support personnel visualize product models, and also let customers visualize retail products in empty space. With the added complexity of tight integration with backend ERP systems.
FEATURED WORK
TECHNOLOGY STACK
Some of the frameworks and tools that our development teams have used recently. A list that grows by the day.
RELATED RESEARCH
Relevant, interesting and current curated research content in the field.
Segment-based Methods for Facial Attribute Detection from Partial Faces
State-of-the-art methods of attribute detection from faces (like those used in Appliqant) almost always assume the presence of a full, unoccluded face. Hence, their performance degrades for partially visible and occluded faces. This paper discusses the use of a deep convolutional neural network-based method that is explicitly designed to perform attribute detection in partially occluded faces. Taking several facial segments and the full face as input, the proposed method takes a data driven approach to determine which attributes are localized in which facial segments.
Multi-Scale Attention with Dense Encoder for Handwritten Mathematical Expression Recognition
Handwritten mathematical expression recognition is a challenging problem due to the complicated two-dimensional structures, ambiguous handwriting input and variant scales of handwritten math symbols. This paper discusses the attention based encoder-decoder model that recognizes mathematical expression images from two-dimensional layouts to one-dimensional LaTeX strings. The encoder is improved by employing densely connected convolutional networks as they can strengthen feature extraction and facilitate gradient propagation especially on a small training set.
DeepTraffic: Driving Fast through Dense Traffic with Deep Reinforcement Learning
This paper presents a micro-traffic simulation (named “DeepTraffic”) where the perception, control, and planning systems for one of the cars are all handled by a single neural network as part of a model-free, off-policy reinforcement learning process. The paper also investigates the crowd-sourced hyperparameter tuning of the policy network that resulted from the first iteration of the DeepTraffic competition where thousands of participants actively searched through the hyperparameter space with the objective of their neural network submission to make it onto the top-10 leaderboard.
Novel Methods for Enhancing the Performance of Genetic Algorithms
This paper discusses the best approaches to take with Genetic algorithms (GA), a branch of evolutionary computing (EC) that mimics the theory of evolution and natural selection, where the technique is based on an heuristic random search. Crossover and mutation are the key to success in genetic algorithms. With the existence of several methods of crossover and mutation operators, this paper attempts to determine which method is best suited to each problem.