Despite the success of deep neural networks (DNNs), state-of-the-art models are too large to deploy on low-resource devices or common server configurations in which multiple models are held in memory. Model compression methods address this limitation by reducing the memory footprint, latency, or energy consumption of a model with minimal impact on accuracy. We focus on the task of reducing the number of learnable variables in the model. In this work we combine ideas from weight hashing and dimensionality reductions resulting in a simple and powerful structured multi-hashing method based on matrix products that allows direct control of model size of any deep network and is trained end-to-end. We demonstrate the strength of our approach by compressing models from the ResNet, EfficientNet, and MobileNet architecture families. Our method allows us to drastically decrease the number of variables while maintaining high accuracy. For instance, by applying our approach to EfficentNet-B4 (16M parameters) we reduce it to the size of B0 (5M parameters), while gaining over 3% in accuracy over B0 baseline. On the commonly used benchmark CIFAR10 we reduce the ResNet32 model by 75% with no loss in quality, and are able to do a 10x compression while still achieving above 90% accuracy.

Shallow depth-of-field is commonly used by photographers to isolate a subject from a distracting background. However, standard cell phone cameras cannot produce such images optically, as their short focal lengths and small apertures capture nearly all-in-focus images. We present a system to computationally synthesize shallow depth-of-field images with a single mobile camera and a single button press. If the image is of a person, we use a person segmentation network to separate the person and their accessories from the background. If available, we also use dense dual-pixel auto-focus hardware, effectively a 2-sample light field with an approximately 1 millimeter baseline, to compute a dense depth map. These two signals are combined and used to render a defocused image. Our system can process a 5.4 megapixel image in 4 seconds on a mobile phone, is fully automatic, and is robust enough to be used by non-experts. The modular nature of our system allows it to degrade naturally in the absence of a dual-pixel sensor or a human subject.

We address the problem of distance metric learning (DML), defined as learning a distance consistent with a notion of semantic similarity. Traditionally, for this problem supervision is expressed in the form of sets of points that follow an ordinal relationship - an anchor point x is similar to a set of positive points Y , and dissimilar to a set of negative points Z, and a loss defined over these distances is minimized. While the specifics of the optimization differ, in this work we collectively call this type of supervision Triplets and all methods that follow this pattern Triplet-Based methods. These methods are challenging to optimize. A main issue is the need for finding informative triplets, which is usually achieved by a variety of tricks such as increasing the batch size, hard or semi-hard triplet mining, etc. Even with these tricks, the convergence rate of such methods is slow. In this paper we propose to optimize the triplet loss on a different space of triplets, consisting of an anchor data point and similar and dissimilar proxy points which are learned as well. These proxies approximate the original data points, so that a triplet loss over the proxies is a tight upper bound of the original loss. This proxy-based loss is empirically better behaved. As a result, the proxy-loss improves on state-of-art results for three standard zero-shot learning datasets, by up to 15% points, while converging three times as fast as other triplet-based losses.

Modern search engines receive large numbers of business related, local aware queries. Such queries are best answered using accurate, up-to-date, business listings, that contain representations of business categories. Creating such listings is a challenging task as businesses often change hands or close down. For businesses with street side locations one can leverage the abundance of street level imagery, such as Google Street View, to automate the process. However, while data is abundant, labeled data is not; the limiting factor is creation of large scale labeled training data. In this work, we utilize an ontology of geographical concepts to automatically propagate business category information and create a large, multi label, training dataset for fine grained storefront classification. Our learner, which is based on the GoogLeNet/Inception Deep Convolutional Network architecture and classifies 208 categories, achieves human level accuracy.

We present Marvin, a system that can search physical objects using a mobile or wearable device. It integrates HOG-based object recognition, SURF-based localization information, automatic speech recognition, and user feedback information with a probabilistic model to recognize the “object of interest” at high accuracy and at interactive speeds. Once the object of interest is recognized, the information that the user is querying, e.g. reviews, options, etc., is displayed on the user's mobile or wearable device. We tested this prototype in a real-world retail store during business hours, with varied degree of background noise and clutter. We show that this multi-modal approach achieves superior recognition accuracy compared to using a vision system alone, especially in cluttered scenes where a vision system would be unable to distinguish which object is of interest to the user without additional input. It is computationally able to scale to large numbers of objects by focusing compute-intensive resources on the objects most likely to be of interest, inferred from user speech and implicit localization information. We present the system architecture, the probabilistic model that integrates the multi-modal information, and empirical results showing the benefits of multi-modal integration.

Question answering (Q&A) communities have been gaining popularity in the past few years. The success of such sites depends mainly on the contribution of a small number of expert users who provide a significant portion of the helpful answers, and so identifying users that have the potential of becoming strong contributers is an important task for owners of such communities.

Moving objects are usually detected by measuring the appearance change from a background model. The background model should adapt to slow changes such as illumination, but detect faster changes caused by moving objects. Particle filters do an excellent task in modeling non parametric distributions as needed for a background model, but may adapt too quickly to the foreground objects. A persistent particle filter is proposed, following bacterial persistence. Bacterial persistence is linked to the random switch of bacteria between two states: A normal growing cell and a dormant but persistent cell. The dormant cells can survive stress such as antibiotics. When a dormant cell switches to a normal status after the stress is over, bacterial growth continues. Similar to bacteria, particles will switch between dormant and active states, where dormant particles will not adapt to the changing environment. A further modification of particle filters allows discontinuous jumps into new parameters enabling foreground objects to join the background when they stop moving. This can also quickly build multi-modal distributions.