In recent years, deep learning models have pushed state-of-the-art accuracies for several machine learning tasks. However, such models require a large amount of (supervised) data for training. While unlabelled data is available in abundance, manually labeling them is very costly. Active learning techniques helps in utilizing unlabelled data which may result in an improved classification model. In this research, we present an active learning algorithm which can help in increasing performance of deep learning models by using large amount of unlabelled data. A novel active learning algorithm, Triplet AL is proposed which uses a triplet network to select samples from an unlabelled data set. Previous active learning methods rely on classification model's final prediction scores as a measure of confidence for an unlabelled sample. We propose a more reliable confidence measure, termed as Top-Two-Margin which is given by the Triplet Network. The proposed algorithm shows improved performance compared to other active learning approaches.

Low cost time-of-flight based depth sensors such as Kinect have opened new avenues for their usage in video surveillance scenarios. RGB-D images obtained from such sensors have shown their utility in improved face recognition capabilities. Generally, existing RGB-D face recognition methods fuse the depth information with RGB information which results in enhanced recognition performance. However, in the real world surveillance scenarios, cameras are placed at a distance too large for low cost depth sensors to capture good quality depth information. Such poor quality depth information may not contribute significantly to face recognition. In this paper, we present a novel representation learning algorithm by learning shared representation of RGB and depth information using a reconstruction based deep neural network. The proposed network, once trained in the offline mode, can generate a shared representation of RGB and depth data using only the RGB image. This feature rich representation is then utilized for face identification. This allows the framework to be used in scenarios where low quality or no depth image is captured. Experiments on multiple real world databases show the effectiveness of the proposed approach.

Software security techniques rely on correct execution by the hardware. Securing hardware components has been challenging due to their complexity and the proportionate attack surface they present during their design, manufacture, deployment, and operation. Recognizing that external communication represents one of the greatest threats to a system's security, this paper introduces the TrustGuard containment architecture. TrustGuard contains malicious and erroneous behavior using a relatively simple and pluggable gatekeeping hardware component called the Sentry. The Sentry bridges a physical gap between the untrusted system and its external interfaces. TrustGuard allows only communication that results from the correct execution of trusted software, thereby preventing the ill effects of actions by malicious hardware or software from leaving the system. The simplicity and pluggability of the Sentry, which is implemented in less than half the lines of code of a simple in-order processor, enables additional measures to secure this root of trust, including formal verification, supervised manufacture, and supply chain diversification with less than a 15% impact on performance.

Deep metric learning algorithms have been utilized to learn discriminative and generalizable models which are effective for classifying unseen classes. In this paper, a novel noise tolerant deep metric learning algorithm is proposed. The proposed method, termed as Density Aware Metric Learning, enforces the model to learn embeddings that are pulled towards the most dense region of the clusters for each class. It is achieved by iteratively shifting the estimate of the center towards the dense region of the cluster thereby leading to faster convergence and higher generalizability. In addition to this, the approach is robust to noisy samples in the training data, often present as outliers. Detailed experiments and analysis on two challenging cross-modal face recognition databases and two popular object recognition databases exhibit the efficacy of the proposed approach. It has superior convergence, requires lesser training time, and yields better accuracies than several popular deep metric learning methods.

Existing face recognition algorithms achieve high recognition performance for frontal face images with good illumination and close proximity to the imaging device. However, most of the existing algorithms fail to perform equally well in surveillance scenarios, where videos are captured across varying resolutions and spectra. In surveillance settings, cameras are usually placed far away from the subjects, thereby resulting in variations across pose, illumination, occlusion, and resolution. Current video datasets used for face recognition are often captured in constrained environments, and thus fail to simulate the real world scenarios. In this paper, we present the FaceSurv database featuring 252 subjects in 460 videos. The proposed dataset contains over 142K face images, spread across videos captured in both visible and near-infrared spectra. Each video contains a group of individuals walking from 36ft towards the imaging device, offering a plethora of challenges common to surveillance settings. Benchmark experimental protocol and baseline results have been reported with state-of-the-art algorithms for face detection and recognition. It is our assertion that the availability of such a challenging database will facilitate the development of robust face recognition systems relevant to real world surveillance scenarios.

Autonomous driving is an active area of research. An important aspect of this problem is recognizing the gestures made by humans, both inside and outside the vehicle. In this paper, we present the Auto-G database that comprises different hand gestures for autonomous driving. The database encompasses several challenges such as occlusion, low resolution, motion blur, illumination variation, extreme pose variations, along with the presence of multiple gestures within a frame. We also propose an end-to-end pipeline for hand detection and gesture recognition. The proposed pipeline achieves a frame gesture recognition accuracy of 90.23% on the proposed Auto-G database.

Automatic parallelization has shown promise in producing scalable multi-threaded programs for multi-core architectures. Most existing automatic techniques parallelize independent loops and insert global synchronization between loop invocations. For programs with many loop invocations, frequent synchronization often becomes the performance bottleneck. Some techniques exploit cross-invocation parallelism to overcome this problem. Using static analysis, they partition iterations among threads to avoid cross-thread dependences. However, this approach may fail if dependence pattern information is not available at compile time. To address this limitation, this work proposes SpecCross--the first automatic parallelization technique to exploit cross-invocation parallelism using speculation. With speculation, iterations from different loop invocations can execute concurrently, and the program synchronizes only on misspeculation. This allows SpecCross to adapt to dependence patterns that only manifest on particular inputs at runtime. Evaluation on eight programs shows that SpecCross achieves a geomean speedup of 3.43x over parallel execution without cross-invocation parallelization.

Face identification from low quality and low resolution Near-Infrared (NIR) face images is a challenging problem. Since surveillance cameras typically acquire images at a large standoff distance, the effective resolution of the face is not large enough to identify the individuals. Moreover for a 24-hour surveillance footage, images in low light and at nighttime are acquired in NIR mode which makes the identification problem even more challenging. We propose an effective method using both hand-crafted and learned features for face identification of low resolution NIR images. We show that learned features contribute considerably to the performance of identification algorithm, and that using both feature level and score level fusion in a hierarchal approach gives good performance. The results demonstrate the effectiveness of the proposed approach on images which are of low quality, low resolution and acquired under challenging illumination conditions in near-infrared mode by surveillance cameras.

Ocular recognition algorithms, including iris matching, have been used in several applications including large scale national ID projects such as India's Aadaar. Deployment of large-scale biometric systems is expected to rely on using multiple devices including mobile devices to ensure widespread adoption of biometric recognition systems. Ocular images captured using mobile devices may have challenges such as uncontrolled illumination, complex background, and geometric distortions. Further, among many enrollees of large scale biometrics program, some may have ocular diseases. One of the most common ocular disease in elderly is cataract. While it is established that iris recognition may be challenging due to ocular diseases, this paper investigates periocular recognition with pre and post cataract surgery images. In this research, we present a mobile periocular database of 145 subjects1. Baseline results also include a framework that achieves over 69% rank-10 accuracy and around 24% genuine accept rate at 1% false accept rate in inter-session experiments.

Microprocessors designed using HW/SW codesign principles, such as Transmeta™ Efficeon™ and the soon-to-ship NVIDIA 64-bit Tegra® K1, use dynamic binary optimization to extract instruction-level parallelism. Many code optimizations are made significantly more effective through the use of alias speculation. The state-of-the-art alias speculation system, SMARQ, provides 40% speedup on average over a system with no alias speculation. This performance, however, comes at the cost of introducing new alias registers and increased power consumption due to new checks for validating speculation. Consequently, improving the efficiency of alias speculation by reducing alias register requirements and rationalizing speculation validation checks is critical for the viability of SMARQ. This paper presents alias coalescing, a novel technique to significantly improve the efficiency of SMARQ through a synergistic combination of compiler and microarchitectural techniques. By using a more compact encoding for memory access ranges for memory instructions, alias coalescing simultaneously reduces the alias register pressure in SMARQ by a geomean of 26.09% and 39.96%, and the dynamic alias checks by 20.73% and 33.87%, across the entire SPEC CINT2006 and SPEC CFP2006 suites respectively.

High-level synthesis (HLS) tools dramatically reduce the nonrecurring engineering cost of creating specialized hardware accelerators. Existing HLS tools are successful in synthesizing efficient accelerators for program kernels with regular memory accesses and simple control flows. For other programs, however, these tools yield poor performance because they invoke computation units for instructions sequentially, without exploiting parallelism. To address this problem, this paper proposes Coarse-Grained Pipelined Accelerators (CGPA), an HLS framework that utilizes coarsegrained pipeline parallelism techniques to synthesize efficient specialized accelerator modules from irregular C/C++ programs without requiring any annotations. Compared to the sequential method, CGPA shows speedups of 3.0x--3.8x for 5 kernels from programs in different domains.

Transient faults are emerging as a critical reliability concern in modern microprocessors. Redundant hardware solutions are commonly deployed to detect transient faults, but they are less flexible and cost-effective than software solutions. However, software solutions are rendered impractical because of high performance overheads. To address this problem, this paper presents Runtime Asynchronous Fault Tolerance via Speculation (RAFT), the fastest transient fault detection technique known to date. Serving as a layer between the application and the underlying platform, RAFT automatically generates two symmetric program instances from a program binary. It detects transient faults in a non-invasive way and exploits high-confidence value speculation to achieve low runtime overhead. Evaluation on a commodity multicore system demonstrates that RAFT delivers a geomean performance overhead of 2.83% on a set of 30 SPEC CPU benchmarks and STAMP benchmarks. Compared with existing transient fault detection techniques, RAFT exhibits the best performance and fault coverage, without requiring any change to the hardware or the software applications.

Sequential programming models express a total program order, of which a partial order must be respected. This inhibits parallelizing tools from extracting scalable performance. Programmer written semantic commutativity assertions provide a natural way of relaxing this partial order, thereby exposing parallelism implicitly in a program. Existing implicit parallel programming models based on semantic commutativity either require additional programming extensions, or have limited expressiveness. This paper presents a generalized semantic commutativity based programming extension, called Commutative Set (COMMSET), and associated compiler technology that enables multiple forms of parallelism. COMMSET expressions are syntactically succinct and enable the programmer to specify commutativity relations between groups of arbitrary structured code blocks. Using only this construct, serializing constraints that inhibit parallelization can be relaxed, independent of any particular parallelization strategy or concurrency control mechanism. COMMSET enables well performing parallelizations in cases where they were inapplicable or non-performing before. By extending eight sequential programs with only 8 annotations per program on average, COMMSET and the associated compiler technology produced a geomean speedup of 5.7x on eight cores compared to 1.5x for the best non-COMMSET parallelization.