While recent advancements in camera-based 3D object detection demonstrate remarkable performance, they require thousands or even millions of human-annotated frames. This requirement significantly inhibits their deployment in various locations and sensor configurations. To address this gap, we propose a performant semi-supervised framework that leverages unlabeled RGB-only driving sequences - data easily collected with cost-effective RGB cameras - to significantly improve temporal, camera-only 3D detectors. We observe that the standard semi-supervised pseudo-labeling paradigm underperforms in this temporal, camera-only setting due to poor 3D localization of pseudo-labels. To address this, we train a single 3D detector to handle RGB sequences both forward and backward in time, then ensemble both its forwards and backwards pseudo-labels for semi-supervised learning. We further improve the pseudo-label quality by leveraging 3D object tracking to infill missing detections and by eschewing simple confidence thresholding in favor of using the auxiliary 2D detection head to filter 3D predictions. Finally, to enable the backbone to learn directly from the unlabeled data itself, we introduce an object-query conditioned masked reconstruction objective. Our framework demonstrates remarkable performance improvement on large-scale autonomous driving datasets nuScenes and nuPlan.

We propose a novel method called Short-/Long-term Hybrid Prediction (SLHP) that predicts short-term and long-term trajectories of surrounding objects while estimating their future influences on an autonomous ego-vehicle for both types of trajectories. Recently, long-term prediction methods based on trajectory sample generation and verification with road environments and/or interactions have been proposed; however, they entail high computational costs because they need to generate and verify multiple trajectory samples for multiple objects. Therefore, they are not appropriate in scenes where short-term prediction is required such as for sudden motions by surrounding objects. In contrast, our SLHP consists of a hybrid prediction based on short-term and long-term trajectory predictors. SLHP provides flexible predictions that are appropriate to scenes caused by objects' motions, road environments, interactions, and so on. In this paper, we apply and evaluate our method to cut-in prediction as a typical prediction task by using the public road dataset that includes various cut-in events. Experimental results show that SLHP achieves a correctness rate of F-measure = 0.86 for cut-in prediction. Additionally, we confirmed the effectiveness of our hybrid prediction method that provides prediction as early as 3.57 s and 4.82 s before the cut-in event for short-term and long-term trajectory predictions, respectively.

Document comprehensibility is one of key factors determining document quality and, in result, user's satisfaction. Relevant web pages are of little utility if they are incomprehensible or impose too much cognitive burden on readers. Traditional measures of text difficulty focus often on syntactic factors of text such as sentence length, word length, syllable count, or they utilize fixed list of common terms. However, document comprehensibility depends on many factors, of which concreteness and the ease of concept visualization are crucial ones. In this paper, we first propose a method for predicting the concreteness of terms using SVM regression. We then extend it to calculating document concreteness level. The experimental results indicate satisfactory accuracy in estimating both term and document concreteness as well as demonstrate positive correlation between the document concreteness and comprehensibility. Our ultimate goal is to enable comprehension-driven search, which will return both relevant and comprehensible results.

This article describes our method used for the 2007 forecasting competition for neural networks and computational intelligence. We have employed the first-order difference of time series for dealing with the seasonality of the monthly data. Since the differencing removes the trend of time series, we have developed a method to estimate the trend. Moreover, we have used the bagging of competitive associative net called CAN2 as a learning predictor, where the CAN2 is for learning an efficient piecewise linear approximation of a nonlinear function, and the bagging for reducing the variance of the prediction.

This article describes the method which we have used for the predictive uncertainty in environmental modelling competition. The method uses competitive associative net called CAN2 embedding piecewise linear approximation scheme. With the scheme, we can naturally estimate piecewise error distribution or heteroscedastic error distribution which may be caused by the noise involved in environmental data. For improving the CAN2 with an efficient batch learning method for reducing empirical (training) error, we introduce ensemble method for more accurate prediction or less prediction (generalization) error. We also introduce multiple K-fold cross-validation for obtaining reliable predictive distribution.

High-speed downlink packet access (HSDPA) is specified in the 3rd generation partnership project (3GPP) to cope with the increasing demand for high-speed Internet-based multimedia services. This paper presents field experimental results of the user throughput performance in HSDPA to clarify the effects of the maximum number of multi-codes, transmit diversity, and advanced user equipment (UE) such as receive diversity and a chip equalizer. The field experimental results show that 15-multi-code UE, transmit diversity, and receive diversity achieve the peak throughput of 6.5, 8.5, and 10.2 Mbps, respectively, at the average moving speed of 30 km/h in an actual multi-path fading propagation channel, and the average received signal-to-interference power ratio (SIR) is 24 dB. Furthermore, the results show that transmit diversity, receive diversity, and the chip equalizer can improve the average throughput to 27.0%, 54.6%, and 12.8%, respectively, compared to the conventional single antenna Rake receiver when the average received SIR is 16.3 dB.

This paper presents coherent adaptive antenna array beamforming (CAAA-BF) combined with maximal ratio combining (MRC) diversity techniques (space diversity, SD, or polarization diversity, PD) and evaluates the interference suppression effect of these methods under multipath fading in the W-CDMA reverse link. Computer simulation results show that even when the number of active users is 150 (note that the capacity normalized by the processing gain, Pg, is 78%), the required transmit signal energy-per-bit-to-background-noise power spectrum density ratio (E/sub b//N/sub 0/) at the average block error rate (BLER) of 10/sup -2/ with CAAA-BF with 2-branch SD reception, i.e., 4 antennas per branch, is decreased by approximately 1.4 dB compared to that with CAAA-BF using 8 antennas. These results indicate that by combining CAAA-BF with diversity techniques, although the interference suppression effect is slightly decreased compared to that with CAAA-BF using the identical number of antennas in an interference-limited channel, due to the broadening beam width, there is a significant additional space diversity effect especially when the number of resolved paths is small.

In wideband direct sequence code division multiple access (W-CDMA), employing an adaptive antenna array is a very promising technique to reduce severe multiple access interference (MAI) from high rate users. A four-antenna pilot symbol-assisted coherent adaptive antenna array diversity (PSA-CAAAD) receiver comprising an adaptive antenna array based on a minimum mean squared error (MMSE) criterion and a RAKE combiner is implemented in preliminary laboratory and field experiments. There are two important design concepts of the PSA-CAAAD receiver. The first is that the adaptive antenna array forms an antenna beam for each resolved propagation path and tracks only slow changes in the directions of arrival (DOAs) and average powers of the desired and interfering user signals. The second is that the RAKE combiner tracks the instantaneous changes in channel conditions and coherently combines the signals of the desired user propagating along the resolved paths to maximize the instantaneous signal-to-interference plus background noise power ratio (SINR). This paper presents, both by laboratory and field experiments, the effectiveness of PSA-CAAAD receiver as a powerful means to reduce severe MAI from high rate users, and that it is more effective than using a space diversity receiver with the same number of antennas in the W-CDMA reverse link.

Gesture is one of the important channels of man-machine interface. In the paper, methods to recognize hand gestures using range images are proposed. Five hand gestures: "come on", "go away", "turn right", "turn left" and "stop" are dealt with. A sequence of velocity vectors and normal vectors of a moving hand are extracted from range images, and gestures are recognized using the features. A recognition system is constructed which utilizes the recognition methods. Experiments to recognize persons' gestures using the recognition system are performed to show the effectiveness of the proposed methods and the recognition system.