Danilo Pau

An innovative, high quality and search window independent motion estimation algorithm and architecture for MPEG-2 encoding

This paper describes an innovative, pipelined, cache-based architecture for a motion estimation coprocessor. The coprocessor is based on a predictive/recursive algorithm whose computational complexity is low and not dependent on the search window. The synergies between the architecture and algorithms features allow a high picture quality, low-area, low-bandwidth, unlimited search window implementation.

A Novel Adaptive Vector Quantization Method For Memory Reduction In Mpeg-2 Hdtv Decoders

A novel method to reduce the external memory needed by an MPEG-2 HDTV decoder architecture is presented and discussed. The total amount of memory is reduced from 96 to 32 Mbits preserving a good picture quality. Furthermore the approach chosen maintains a low hardware complexity and increases less than 5%, in 0.35 /spl mu/m technology, the total decoder silicon area with respect to the standard decoder.

An innovative, programmable architecture for ultra-low power motion estimation in reduced memory MPEG-4 encoder

This paper describes an ultra-low power, cache based, and programmable motion estimator with memory reduction for MPEG-4 video encoding. It exploits a low complexity motion estimation algorithm, achieving a quality comparable to the full search approach with only 1% of the computation and the power consumption.

Toward Compressed 3D Descriptors

Visual search for mobile devices relies on transmitting wirelessly a compact representation of the query image, generally in the form of feature descriptors, to a remote server. Descriptors are therefore compressed, so as to reduce bandwidth occupancy and network latency. Given the impressive pace of growth of 3D video technology, we foresee 3D visual search applications for the mobile and the robotic market to become a reality. Accordingly, our work proposes a study on compressed 3D descriptors, a fundamental building block for such prospective applications. Based on analysis of several compression approaches, we develop and assess different schemes to achieve a compact version of a state-of-the-art 3D descriptor. Through experiments on a vast dataset we demonstrate the ability to achieve compression rates as high as 98% with a negligible loss in 3D visual search performance.

Reduced memory region based deep Convolutional Neural Network detection

Accurate pedestrian detection has a primary role in automotive safety: for example, by issuing warnings to the driver or acting actively on cars brakes, it helps decreasing the probability of injuries and human fatalities. In order to achieve very high accuracy, recent pedestrian detectors have been based on Convolutional Neural Networks (CNN). Unfortunately, such approaches require vast amounts of computational power and memory, preventing efficient implementations on embedded systems. This work proposes a CNN-based detector, adapting a general-purpose convolutional network to the task at hand. By thoroughly analyzing and optimizing each step of the detection pipeline, we develop an architecture that outperforms methods based on traditional image features and achieves an accuracy close to the state-of-the-art while having low computational complexity. Furthermore, the model is compressed in order to fit the tight constrains of low power devices with a limited amount of embedded memory available. This paper makes two main contributions: (1) it proves that a region based deep neural network can be finely tuned to achieve adequate accuracy for pedestrian detection (2) it achieves a very low memory usage without reducing detection accuracy on the Caltech Pedestrian dataset.

Analysis of Compact Features for RGB-D Visual Search

Anticipating the oncoming integration of depth sensing into mobile devices, we experimentally compare different compact features for representing RGB-D images in mobile visual search. Experiments on 3 state-of-the-art datasets, addressing both category and instance recognition, show how Deep Features provided by Convolutional Neural Networks better represent appearance information, whereas shape is more effectively encoded through Kernel Descriptors. Moreover, our evaluation suggests that learning to weight the relative contribution of depth and appearance is key to deploy effectively depth sensing in forthcoming mobile visual search scenarios.

RGB-D Visual Search with Compact Binary Codes

As integration of depth sensing into mobile devices is likely forthcoming, we investigate on merging appearance and shape information for mobile visual search. Accordingly, we propose an RGB-D search engine architecture that can attain high recognition rates with peculiarly moderate bandwidth requirements. Our experiments include a comparison to the CDVS (Compact Descriptors for Visual Search) pipeline, candidate to become part of the MPEG-7 standard, and contribute to elucidate on the merits and limitations of joint deployment of depth and color in mobile visual search.

Complexity and Accuracy of Hand-Crafted Detection Methods Compared to Convolutional Neural Networks

Even though Convolutional Neural Networks have had the best accuracy in the last few years, they have a price in term of computational complexity and memory footprint, due to a large number of multiply-accumulate operations and model parameters. For embedded systems, this complexity severely limits the opportunities to reduce power consumption, which is dominated by memory read and write operations. Anticipating the oncoming integration into intelligent sensor devices, we compare hand-crafted features for the detection of a limited number of objects against some typical convolutional neural network architectures. Experiments on some state-of-the-art datasets, addressing detection tasks, show that for some problems the increased complexity of neural networks is not reflected by a large increase in accuracy. Moreover, our analysis suggests that for embedded devices hand-crafted features are still competitive in terms of accuracy/complexity trade-offs.

Accurate omnidirectional multi-camera embedded structure from motion

Trajectory estimation and 3D scene reconstruction from multiple cameras (also referred as Structure from Motion, SfM) will have a central role in the future of automotive industry. Typical appliance fields will be: autonomous navigation/guidance, collisions avoidance against static or moving objects (in particular pedestrians), parking assisted maneuvers and many more. The work exposed in this paper had mainly two different goals: (1) to describe the implementation of a real time embedded SfM modular pipeline featuring a dedicated optimized HW/SW system partitioning. It included also nonlinear optimizations such as local and global bundle adjustment at different stages of the pipeline; (2) to demonstrate quantitatively its performances on a synthetic test space specifically designed for its characterization. In order to make the system reliable and effective, providing the driver or the autonomous vehicle with a prompt response, the data rates and low latency of the 5G communication systems appear to make this choice the most promising communication solution.

The Orlando Project: A 28 nm FD-SOI Low Memory Embedded Neural Network ASIC

The recent success of neural networks in various computer vision tasks open the possibility to add visual intelligence to mobile and wearable devices; however, the stringent power requirements are unsuitable for networks run on embedded CPUs or GPUs. To address such challenges, STMicroelectronics developed the Orlando Project, a new and low power architecture for convolutional neural network acceleration suited for wearable devices. An important contribution to the energy usage is the storage and access to the neural network parameters. In this paper, we show that with adequate model compression schemes based on weight quantization and pruning, a whole AlexNet network can fit in the local memory of an embedded processor, thus avoiding additional system complexity and energy usage, with no or low impact on the accuracy of the network. Moreover, the compression methods work well across different tasks, e.g. image classification and object detection.