Search for: All records

This paper presents an approach to detect out-of-context (OOC) objects in an image. Given an image with a set of objects, our goal is to determine if an object is inconsistent with the scene context and detect the OOC object with a bounding box. In this work, we consider commonly explored contextual relations such as co-occurrence relations, the relative size of an object with respect to other objects, and the position of the object in the scene. We posit that contextual cues are useful to determine object labels for in-context objects and inconsistent context cues are detrimental to determining object labels for out-of-context objects. To realize this hypothesis, we propose a graph contextual reasoning network (GCRN) to detect OOC objects. GCRN consists of two separate graphs to predict object labels based on the contextual cues in the image: 1) a representation graph to learn object features based on the neighboring objects and 2) a context graph to explicitly capture contextual cues from the neighboring objects. GCRN explicitly captures the contextual cues to improve the detection of in-context objects and identify objects that violate contextual relations. In order to evaluate our approach, we create a large-scale dataset by adding OOC objectmore »instances to the COCO images. We also evaluate on recent OCD benchmark. Our results show that GCRN outperforms competitive baselines in detecting OOC objects and correctly detecting in-context objects.« less

This paper presents an approach to detect out-of-context (OOC) objects in an image. Given an image with a set of objects, our goal is to determine if an object is inconsistent with the scene context and detect the OOC object with a bounding box. In this work, we consider commonly explored contextual relations such as co-occurrence relations, the relative size of an object with respect to other objects, and the position of the object in the scene. We posit that contextual cues are useful to determine object labels for in-context objects and inconsistent context cues are detrimental to determining object labels for out-of-context objects. To realize this hypothesis, we propose a graph contextual reasoning network (GCRN) to detect OOC objects. GCRN consists of two separate graphs to predict object labels based on the contextual cues in the image: 1) a representation graph to learn object features based on the neighboring objects and 2) a context graph to explicitly capture contextual cues from the neighboring objects. GCRN explicitly captures the contextual cues to improve the detection of in-context objects and identify objects that violate contextual relations. In order to evaluate our approach, we create a large-scale dataset by adding OOC objectmore »instances to the COCO images. We also evaluate on recent OCD benchmark. Our results show that GCRN outperforms competitive baselines in detecting OOC objects and correctly detecting in-context objects.« less

This paper presents acoustic propulsion in air by synthesis jets produced by ultrasounds. Various ultrasonic air-borne propellers have been fabricated on 0.37-mm-thick commercial card piezoelectric speakers (APS2513S-T-R, 25.2 × 16.6 × 0.37 mm3 in size), and studied, with the propulsion force measured through a precision weight scale, as the orifice size, thickness, spacing between orifices, and number (in the orifice array) are varied. Also varied is the orifice depth profile, as the fabrication processes for the orifices produce varying profiles. Strongest acoustic propulsion of 5.4 mg is obtained at 66 kHz (far beyond audible range) with 14 × 14 orifice array made on a 0.1-mm-thick polyester plate (resulting in a propeller of 25.2 × 16.6 × 1.37 mm3 in volume and 500 mg in weight). The acoustic propulsion force, though 93 times less than the propeller weight, is capable of making the propeller jump and move laterally.

Deep neural networks (DNNs) have achieved near-human level accuracy on many datasets across different domains. But they are known to produce incorrect predictions with high confidence on inputs far from the training distribution. This challenge of lack of calibration of DNNs has limited the adoption of deep learning models in high-assurance systems such as autonomous driving, air traffic management, cybersecurity, and medical diagnosis. The problem of detecting when an input is outside the training distribution of a machine learning model, and hence, its prediction on this input cannot be trusted, has received significant attention recently. Several techniques based on statistical, geometric, topological, or relational signatures have been developed to detect the out-of-distribution (OOD) or novel inputs. In this paper, we present a runtime monitor based on predictive processing and dual process theory. We posit that the bottom-up deep neural networks can be monitored using top-down context models comprising two layers. The first layer is a feature density model that learns the joint distribution of the original DNN’s inputs, outputs, and the model’s explanation for its decisions. The second layer is a graph Markov neural network that captures an even broader context. We demonstrate the efficacy of our monitoring architecture inmore »recognizing out-of-distribution and out-of-context inputs on the image classification and object detection tasks.« less

Security is a well-known function to any transmission operator and system planner. As the world is moving toward the decarbonization of the power industry, it is more complicated for the system operators to maintain an acceptable level of security in the power system operation. More large-scale wind farms are being incorporated into the grid, and thus, the voltage stability concern is increasing. In practice, several contingencies are imagined by the system operators to assess the reliability of the grid. Since voltage stability is one of the major menaces that can trigger voltage instability in a power system, this paper is attempting to present to the transmission system planners and operators a dedicated methodology to facilitate the incorporation of large-scale wind farms into a transmission grid under high penetration of wind power. the stability of a wind-dominated power system is discussed based on Q-V and P-V methodologies and some N-1 contingencies with the Remedial Action Schemes (RAS). Furthermore, a methodology to rank the worst contingencies and to predict the voltage collapse during the highest wind penetration level is presented. Simulations have been, extensively, carried out to examine the methodology and have provided valuable information about the static security of the wind-dominatedmore »power system. The results can be used by the transmission system operator to anticipate voltage instability or voltage collapse in the power system during high wind penetration levels.« less