Designing efficient neural networks for embedded devices is a critical challenge, particularly in applications requiring real-time performance, such as aerial imaging with drones and UAVs for emergency responses. In this work, we introduce TakuNet, a novel light-weight architecture which employs techniques such as depth-wise convolutions and an early downsampling stem to reduce computational complexity while maintaining high accuracy. It leverages dense connections for fast convergence during training and uses 16-bit floating-point precision for optimization on embedded hardware accelerators. Experimental evaluation on two public datasets shows that TakuNet achieves near-state-of-the-art accuracy in classifying aerial images of emergency situations, despite its minimal parameter count. Real-world tests on embedded devices, namely Jetson Orin Nano and Raspberry Pi, confirm TakuNet's efficiency, achieving more than 650 fps on the 15W Jetson board, making it suitable for real-time AI processing on resource-constrained platforms and advancing the applicability of drones in emergency scenarios. The code and implementation details are publicly released.
Modern applications, such as autonomous vehicles, require deploying deep learning algorithms on resource-constrained edge devices for real-time image and video processing. However, there is limited understanding of the efficiency and performance of various object detection models on these devices. In this paper, we evaluate state-of-the-art object detection models, including YOLOv8 (Nano, Small, Medium), EfficientDet Lite (Lite0, Lite1, Lite2), and SSD (SSD MobileNet V1, SSDLite MobileDet). We deployed these models on popular edge devices like the Raspberry Pi 3, 4, and 5 with/without TPU accelerators, and Jetson Orin Nano, collecting key performance metrics such as energy consumption, inference time, and Mean Average Precision (mAP). Our findings highlight that lower mAP models such as SSD MobileNet V1 are more energy-efficient and faster in inference, whereas higher mAP models like YOLOv8 Medium generally consume more energy and have slower inference, though with exceptions when accelerators like TPUs are used. Among the edge devices, Jetson Orin Nano stands out as the fastest and most energy-efficient option for request handling, despite having the highest idle energy consumption. These results emphasize the need to balance accuracy, speed, and energy efficiency when deploying deep learning models on edge devices, offering valuable guidance for practitioners and researchers selecting models and devices for their applications.
Sound event localization and detection (SELD) is critical for various real-world applications, including smart monitoring and Internet of Things (IoT) systems. Although deep neural networks (DNNs) represent the state-of-the-art approach for SELD, their significant computational complexity and model sizes present challenges for deployment on resource-constrained edge devices, especially under real-time conditions. Despite the growing need for real-time SELD, research in this area remains limited. In this paper, we investigate the unique challenges of deploying SELD systems for real-world, real-time applications by performing extensive experiments on a commercially available Raspberry Pi 3 edge device. Our findings reveal two critical, often overlooked considerations: the high computational cost of feature extraction and the performance degradation associated with low-latency, real-time inference. This paper provides valuable insights and considerations for future work toward developing more efficient and robust real-time SELD systems
The widespread usage of cars and other large, heavy vehicles necessitates the development of an effective parking infrastructure. Additionally, algorithms for detection and recognition of number plates are often used to identify automobiles all around the world where standardized plate sizes and fonts are enforced, making recognition an effortless task. As a result, both kinds of data can be combined to develop an intelligent parking system focuses on the technology of Automatic Number Plate Recognition (ANPR). Retrieving characters from an inputted number plate image is the sole purpose of ANPR which is a costly procedure. In this article, we propose Chaurah, a minimal cost ANPR system that relies on a Raspberry Pi 3 that was specifically created for parking facilities. The system employs a dual-stage methodology, with the first stage being an ANPR system which makes use of two convolutional neural networks (CNNs). The primary locates and recognises license plates from a vehicle image, while the secondary performs Optical Character Recognition (OCR) to identify individualized numbers from the number plate. An application built with Flutter and Firebase for database administration and license plate record comparison makes up the second component of the overall solution. The application also acts as an user-interface for the billing mechanism based on parking time duration resulting in an all-encompassing software deployment of the study.
The paper proposes a novel feature extraction approach for FMCW radar systems in the field of short-range gesture sensing. A light-weight processing is proposed which reduces a series of 3D radar data cubes to four 1D time signals containing information about range, azimuth angle, elevation angle and magnitude. The processing is entirely performed in the time domain without using any Fourier transformation and enables the training of a deep neural network directly on the raw time domain data. It is shown experimentally on real world data, that the proposed processing retains the same expressive power as conventional radar processing to range-, Doppler- and angle-spectrograms. Further, the computational complexity is significantly reduced which makes it perfectly suitable for embedded devices. The system is able to recognize ten different gestures with an accuracy of about 95% and is running in real time on a Raspberry Pi 3 B. The delay between end of gesture and prediction is only 150 ms.
Visual Place Recognition (VPR) has been a subject of significant research over the last 15 to 20 years. VPR is a fundamental task for autonomous navigation as it enables self-localization within an environment. Although robots are often equipped with resource-constrained hardware, the computational requirements of and effects on VPR techniques have received little attention. In this work, we present a hardware-focused benchmark evaluation of a number of state-of-the-art VPR techniques on public datasets. We consider popular single board computers, including ODroid, UP and Raspberry Pi 3, in addition to a commodity desktop and laptop for reference. We present our analysis based on several key metrics, including place-matching accuracy, image encoding time, descriptor matching time and memory needs. Key questions addressed include: (1) How does the performance accuracy of a VPR technique change with processor architecture? (2) How does power consumption vary for different VPR techniques and embedded platforms? (3) How much does descriptor size matter in comparison to today's embedded platforms' storage? (4) How does the performance of a high-end platform relate to an on-board low-end embedded platform for VPR? The extensive analysis and results in this work serve not only as a benchmark for the VPR community, but also provide useful insights for real-world adoption of VPR applications.
Optical Character Recognition (OCR) has many real world applications. The existing methods normally detect where the characters are, and then recognize the character for each detected location. Thus the accuracy of characters recognition is impacted by the performance of characters detection. In this paper, we propose a method for recognizing characters without detecting the location of each character. This is done by converting the OCR task into an image captioning task. One advantage of the proposed method is that the labeled bounding boxes for the characters are not needed during training. The experimental results show the proposed method outperforms the existing methods on both the license plate recognition and the watermeter character recognition tasks. The proposed method is also deployed into a low-power (300mW) CNN accelerator chip connected to a Raspberry Pi 3 for on-device applications.
Lower limb prosthesis can benefit from embedded systems capable of applying computer vision techniques to enhance autonomous control and context awareness for intelligent decision making. In order to fill in the gap of current literature of autonomous systems for prosthetic legs employing computer vision methods, we evaluate the performance capabilities of two off-the-shelf platforms, the Jetson TX2 and Raspberry Pi 3, by assessing their CPU load, memory usage, run time and classification accuracy for different image sizes and widespread computer vision algorithms. We make use of a dataset that we collected for terrain recognition using images from a camera mounted on a leg, which would enable context-awareness for lower limb prosthesis. We show that, given reasonably large images and an appropriate frame selection method, it is possible to identify the terrain that a subject is going to step on, with the possibility of reconstructing the surface and obtaining its inclination. This is part of a proposed system equipped with an embedded camera and inertial measurement unit to recognize different types of terrain and estimate the slope of the surface in front of the subject.
Plankton provide the foundation for life on earth. To advance our understanding of the marine ecosystem, for scientific, commercial and survival purposes, more in situ continuous monitoring and analysis of plankton is required. Cost, complexity, power and data communication demands are barriers to widespread deployment of in situ plankton microscopes. We address these barriers by building and characterizing a lensless microscope with a data pipeline optimized for the Raspberry Pi 3. The pipeline records 1080p video of multiple plankton swimming in a sample well while simultaneously detecting, tracking and selecting salient cropped images for classification @ 5.1 frames per second. Thirteen machine learning classifiers and combinations of nine sets of features are evaluated on nine plankton classes, optimized for speed (F1=0.74 @ 1 msec. per image prediction) and accuracy (F1=0.81 @ .80 sec.). System performance results confirm that performing the entire data pipeline from image capture to classification is possible on a low-cost open-source embedded computer.
Hyperdimensional computing (HD) is an emerging paradigm for machine learning based on the evidence that the brain computes on high-dimensional, distributed, representations of data. The main operation of HD is encoding, which transfers the input data to hyperspace by mapping each input feature to a hypervector, accompanied by so-called bundling procedure that simply adds up the hypervectors to realize encoding hypervector. Although the operations of HD are highly parallelizable, the massive number of operations hampers the efficiency of HD in embedded domain. In this paper, we propose SHEARer, an algorithm-hardware co-optimization to improve the performance and energy consumption of HD computing. We gain insight from a prudent scheme of approximating the hypervectors that, thanks to inherent error resiliency of HD, has minimal impact on accuracy while provides high prospect for hardware optimization. In contrast to previous works that generate the encoding hypervectors in full precision and then ex-post quantizing, we compute the encoding hypervectors in an approximate manner that saves a significant amount of resources yet affords high accuracy. We also propose a novel FPGA implementation that achieves striking performance through massive parallelism with low power consumption. Moreover, we develop a software framework that enables training HD models by emulating the proposed approximate encodings. The FPGA implementation of SHEARer achieves an average throughput boost of 104,904x (15.7x) and energy savings of up to 56,044x (301x) compared to state-of-the-art encoding methods implemented on Raspberry Pi 3 (GeForce GTX 1080 Ti) using practical machine learning datasets.
Predicting depth from a single image is an attractive research topic since it provides one more dimension of information to enable machines to better perceive the world. Recently, deep learning has emerged as an effective approach to monocular depth estimation. As obtaining labeled data is costly, there is a recent trend to move from supervised learning to unsupervised learning to obtain monocular depth. However, most unsupervised learning methods capable of achieving high depth prediction accuracy will require a deep network architecture which will be too heavy and complex to run on embedded devices with limited storage and memory spaces. To address this issue, we propose a new powerful network with a recurrent module to achieve the capability of a deep network while at the same time maintaining an extremely lightweight size for real-time high performance unsupervised monocular depth prediction from video sequences. Besides, a novel efficient upsample block is proposed to fuse the features from the associated encoder layer and recover the spatial size of features with the small number of model parameters. We validate the effectiveness of our approach via extensive experiments on the KITTI dataset. Our new model can run at a speed of about 110 frames per second (fps) on a single GPU, 37 fps on a single CPU, and 2 fps on a Raspberry Pi 3. Moreover, it achieves higher depth accuracy with nearly 33 times fewer model parameters than state-of-the-art models. To the best of our knowledge, this work is the first extremely lightweight neural network trained on monocular video sequences for real-time unsupervised monocular depth estimation, which opens up the possibility of implementing deep learning-based real-time unsupervised monocular depth prediction on low-cost embedded devices.
Real-time apple detection in orchards is one of the most effective ways of estimating apple yields, which helps in managing apple supplies more effectively. Traditional detection methods used highly computational machine learning algorithms with intensive hardware set up, which are not suitable for infield real-time apple detection due to their weight and power constraints. In this study, a real-time embedded solution inspired from "Edge AI" is proposed for apple detection with the implementation of YOLOv3-tiny algorithm on various embedded platforms such as Raspberry Pi 3 B+ in combination with Intel Movidius Neural Computing Stick (NCS), Nvidia's Jetson Nano and Jetson AGX Xavier. Data set for training were compiled using acquired images during field survey of apple orchard situated in the north region of Italy, and images used for testing were taken from widely used google data set by filtering out the images containing apples in different scenes to ensure the robustness of the algorithm. The proposed study adapts YOLOv3-tiny architecture to detect small objects. It shows the feasibility of deployment of the customized model on cheap and power-efficient embedded hardware without compromising mean average detection accuracy (83.64%) and achieved frame rate up to 30 fps even for the difficult scenarios such as overlapping apples, complex background, less exposure of apple due to leaves and branches. Furthermore, the proposed embedded solution can be deployed on the unmanned ground vehicles to detect, count, and measure the size of the apples in real-time to help the farmers and agronomists in their decision making and management skills.
With the rapid development of technology, automobiles have become an essential asset in our day-to-day lives. One of the more important researches is Traffic Signs Recognition (TSR) systems. This paper describes an approach for efficiently detecting and recognizing traffic signs in real-time, taking into account the various weather, illumination and visibility challenges through the means of transfer learning. We tackle the traffic sign detection problem using the state-of-the-art of multi-object detection systems such as Faster Recurrent Convolutional Neural Networks (F-RCNN) and Single Shot Multi- Box Detector (SSD) combined with various feature extractors such as MobileNet v1 and Inception v2, and also Tiny-YOLOv2. However, the focus of this paper is going to be F-RCNN Inception v2 and Tiny YOLO v2 as they achieved the best results. The aforementioned models were fine-tuned on the German Traffic Signs Detection Benchmark (GTSDB) dataset. These models were tested on the host PC as well as Raspberry Pi 3 Model B+ and the TASS PreScan simulation. We will discuss the results of all the models in the conclusion section.
Deepspeech was very useful for development IoT devices that need voice recognition. One of the voice recognition systems is deepspeech from Mozilla. Deepspeech is an open-source voice recognition that was using a neural network to convert speech spectrogram into a text transcript. This paper shows the implementation process of speech recognition on a low-end computational device. Development of English-language speech recognition that has many datasets become a good point for starting. The model that used results from pre-trained model that provide by each version of deepspeech, without change of the model that already released, furthermore the benefit of using raspberry pi as a media end-to-end speech recognition device become a good thing, user can change and modify of the speech recognition, and also deepspeech can be standalone device without need continuously internet connection to process speech recognition, and even this paper show the power of Tensorflow Lite can make a significant difference on inference by deepspeech rather than using Tensorflow non-Lite.This paper shows the experiment using Deepspeech version 0.1.0, 0.1.1, and 0.6.0, and there is some improvement on Deepspeech version 0.6.0, faster while processing speech-to-text on old hardware raspberry pi 3 b+.
With the advancements in high volume, low precision computational technology and applied research on cognitive artificially intelligent heuristic systems, machine learning solutions through neural networks with real-time learning has seen an immense interest in the research community as well the industry. This paper involves research, development and experimental analysis of a neural network implemented on a robot with an arm through which evolves to learn to walk in a straight line or as required. The neural network learns using the algorithms of Gradient Descent and Backpropagation. Both the implementation and training of the neural network is done locally on the robot on a raspberry pi 3 so that its learning process is completely independent. The neural network is first tested on a custom simulator developed on MATLAB and then implemented on the raspberry computer. Data at each generation of the evolving network is stored, and analysis both mathematical and graphical is done on the data. Impact of factors like the learning rate and error tolerance on the learning process and final output is analyzed.
With the emergence of onboard vision processing for areas such as the internet of things (IoT), edge computing and autonomous robots, there is increasing demand for computationally efficient convolutional neural network (CNN) models to perform real-time object detection on resource constraints hardware devices. Tiny-YOLO is generally considered as one of the faster object detectors for low-end devices and is the basis for our work. Our experiments on this network have shown that Tiny-YOLO can achieve 0.14 frames per second(FPS) on the Raspberry Pi 3 B, which is too slow for soccer playing autonomous humanoid robots detecting goal and ball objects. In this paper we propose an adaptation to the YOLO CNN model named xYOLO, that can achieve object detection at a speed of 9.66 FPS on the Raspberry Pi 3 B. This is achieved by trading an acceptable amount of accuracy, making the network approximately 70 times faster than Tiny-YOLO. Greater inference speed-ups were also achieved on a desktop CPU and GPU. Additionally we contribute an annotated Darknet dataset for goal and ball detection.
Decision-tree-based ensemble classification methods (DTEMs) are a prevalent tool for supervised anomaly detection. However, due to the continued growth of datasets, DTEMs result in increasing drawbacks such as growing memory footprints, longer training times, and slower classification latencies at lower throughput. In this paper, we present, design, and evaluate RADE - a DTEM-based anomaly detection framework that augments standard DTEM classifiers and alleviates these drawbacks by relying on two observations: (1) we find that a small (coarse-grained) DTEM model is sufficient to classify the majority of the classification queries correctly, such that a classification is valid only if its corresponding confidence level is greater than or equal to a predetermined classification confidence threshold; (2) we find that in these fewer harder cases where our coarse-grained DTEM model results in insufficient confidence in its classification, we can improve it by forwarding the classification query to one of expert DTEM (fine-grained) models, which is explicitly trained for that particular case. We implement RADE in Python based on scikit-learn and evaluate it over different DTEM methods: RF, XGBoost, AdaBoost, GBDT and LightGBM, and over three publicly available datasets. Our evaluation over both a strong AWS EC2 instance and a Raspberry Pi 3 device indicates that RADE offers competitive and often superior anomaly detection capabilities as compared to standard DTEM methods, while significantly improving memory footprint (by up to 5.46x), training-time (by up to 17.2x), and classification latency (by up to 31.2x).
Face detection, as a fundamental technology for various applications, is always deployed on edge devices which have limited memory storage and low computing power. This paper introduces a Light and Fast Face Detector (LFFD) for edge devices. The proposed method is anchor-free and belongs to the one-stage category. Specifically, we rethink the importance of receptive field (RF) and effective receptive field (ERF) in the background of face detection. Essentially, the RFs of neurons in a certain layer are distributed regularly in the input image and theses RFs are natural "anchors". Combining RF "anchors" and appropriate RF strides, the proposed method can detect a large range of continuous face scales with 100% coverage in theory. The insightful understanding of relations between ERF and face scales motivates an efficient backbone for one-stage detection. The backbone is characterized by eight detection branches and common layers, resulting in efficient computation. Comprehensive and extensive experiments on popular benchmarks: WIDER FACE and FDDB are conducted. A new evaluation schema is proposed for application-oriented scenarios. Under the new schema, the proposed method can achieve superior accuracy (WIDER FACE Val/Test -- Easy: 0.910/0.896, Medium: 0.881/0.865, Hard: 0.780/0.770; FDDB -- discontinuous: 0.973, continuous: 0.724). Multiple hardware platforms are introduced to evaluate the running efficiency. The proposed method can obtain fast inference speed (NVIDIA TITAN Xp: 131.45 FPS at 640x480; NVIDIA TX2: 136.99 PFS at 160x120; Raspberry Pi 3 Model B+: 8.44 FPS at 160x120) with model size of 9 MB.
We propose a new deep learning based framework to identify pedestrians, and caution distracted drivers, in an effort to prevent the loss of life and property. This framework uses two Convolutional Neural Networks (CNN), one which detects pedestrians and the second which predicts the onset of drowsiness in a driver, is implemented on a Raspberry Pi 3 Model B+, shows great promise. The algorithm for implementing such a low-cost, low-compute model is presented and the results discussed.
Unsupervised depth estimation from a single image is a very attractive technique with several implications in robotic, autonomous navigation, augmented reality and so on. This topic represents a very challenging task and the advent of deep learning enabled to tackle this problem with excellent results. However, these architectures are extremely deep and complex. Thus, real-time performance can be achieved only by leveraging power-hungry GPUs that do not allow to infer depth maps in application fields characterized by low-power constraints. To tackle this issue, in this paper we propose a novel architecture capable to quickly infer an accurate depth map on a CPU, even of an embedded system, using a pyramid of features extracted from a single input image. Similarly to state-of-the-art, we train our network in an unsupervised manner casting depth estimation as an image reconstruction problem. Extensive experimental results on the KITTI dataset show that compared to the top performing approach our network has similar accuracy but a much lower complexity (about 6% of parameters) enabling to infer a depth map for a KITTI image in about 1.7 s on the Raspberry Pi 3 and at more than 8 Hz on a standard CPU. Moreover, by trading accuracy for efficiency, our network allows to infer maps at about 2 Hz and 40 Hz respectively, still being more accurate than most state-of-the-art slower methods. To the best of our knowledge, it is the first method enabling such performance on CPUs paving the way for effective deployment of unsupervised monocular depth estimation even on embedded systems.