Autonomous driving requires the model to perceive the environment and (re)act within a low latency for safety. While past works ignore the inevitable changes in the environment after processing, streaming perception is proposed to jointly evaluate the latency and accuracy into a single metric for video online perception. In this paper, instead of searching trade-offs between accuracy and speed like previous works, we point out that endowing real-time models with the ability to predict the future is the key to dealing with this problem. We build a simple and effective framework for streaming perception. It equips a novel DualFlow Perception module (DFP), which includes dynamic and static flows to capture the moving trend and basic detection feature for streaming prediction. Further, we introduce a Trend-Aware Loss (TAL) combined with a trend factor to generate adaptive weights for objects with different moving speeds. Our simple method achieves competitive performance on Argoverse-HD dataset and improves the AP by 4.9% compared to the strong baseline, validating its effectiveness. Our code will be made available at https://github.com/yancie-yjr/StreamYOLO.
We present DINO (\textbf{D}ETR with \textbf{I}mproved de\textbf{N}oising anch\textbf{O}r boxes), a state-of-the-art end-to-end object detector. % in this paper. DINO improves over previous DETR-like models in performance and efficiency by using a contrastive way for denoising training, a mixed query selection method for anchor initialization, and a look forward twice scheme for box prediction. DINO achieves $49.4$AP in $12$ epochs and $51.3$AP in $24$ epochs on COCO with a ResNet-50 backbone and multi-scale features, yielding a significant improvement of $\textbf{+6.0}$\textbf{AP} and $\textbf{+2.7}$\textbf{AP}, respectively, compared to DN-DETR, the previous best DETR-like model. DINO scales well in both model size and data size. Without bells and whistles, after pre-training on the Objects365 dataset with a SwinL backbone, DINO obtains the best results on both COCO \texttt{val2017} ($\textbf{63.2}$\textbf{AP}) and \texttt{test-dev} (\textbf{$\textbf{63.3}$AP}). Compared to other models on the leaderboard, DINO significantly reduces its model size and pre-training data size while achieving better results. Our code will be available at \url{https://github.com/IDEACVR/DINO}.
Deploying deep neural networks~(DNNs) on edge devices provides efficient and effective solutions for the real-world tasks. Edge devices have been used for collecting a large volume of data efficiently in different domains. DNNs have been an effective tool for data processing and analysis. However, designing DNNs on edge devices is challenging due to the limited computational resources and memory. To tackle this challenge, we demonstrate Object Detection System for Edge Devices~(BED) on the MAX78000 DNN accelerator. It integrates on-device DNN inference with a camera and an LCD display for image acquisition and detection exhibition, respectively. BED is a concise, effective and detailed solution, including model training, quantization, synthesis and deployment. The entire repository is open-sourced on Github, including a Graphical User Interface~(GUI) for on-chip debugging. Experiment results indicate that BED can produce accurate detection with a 300-KB tiny DNN model, which takes only 91.9 ms of inference time and 1.845 mJ of energy. The real-time detection is available at YouTube.
Now a days, UAVs such as drones are greatly used for various purposes like that of capturing and target detection from ariel imagery etc. Easy access of these small ariel vehicles to public can cause serious security threats. For instance, critical places may be monitored by spies blended in public using drones. Study in hand proposes an improved and efficient Deep Learning based autonomous system which can detect and track very small drones with great precision. The proposed system consists of a custom deep learning model Tiny YOLOv3, one of the flavors of very fast object detection model You Look Only Once (YOLO) is built and used for detection. The object detection algorithm will efficiently the detect the drones. The proposed architecture has shown significantly better performance as compared to the previous YOLO version. The improvement is observed in the terms of resource usage and time complexity. The performance is measured using the metrics of recall and precision that are 93% and 91% respectively.
The "Roaring 20s" of visual recognition began with the introduction of Vision Transformers (ViTs), which quickly superseded ConvNets as the state-of-the-art image classification model. A vanilla ViT, on the other hand, faces difficulties when applied to general computer vision tasks such as object detection and semantic segmentation. It is the hierarchical Transformers (e.g., Swin Transformers) that reintroduced several ConvNet priors, making Transformers practically viable as a generic vision backbone and demonstrating remarkable performance on a wide variety of vision tasks. However, the effectiveness of such hybrid approaches is still largely credited to the intrinsic superiority of Transformers, rather than the inherent inductive biases of convolutions. In this work, we reexamine the design spaces and test the limits of what a pure ConvNet can achieve. We gradually "modernize" a standard ResNet toward the design of a vision Transformer, and discover several key components that contribute to the performance difference along the way. The outcome of this exploration is a family of pure ConvNet models dubbed ConvNeXt. Constructed entirely from standard ConvNet modules, ConvNeXts compete favorably with Transformers in terms of accuracy and scalability, achieving 87.8% ImageNet top-1 accuracy and outperforming Swin Transformers on COCO detection and ADE20K segmentation, while maintaining the simplicity and efficiency of standard ConvNets.
Advanced wearable devices are increasingly incorporating high-resolution multi-camera systems. As state-of-the-art neural networks for processing the resulting image data are computationally demanding, there has been growing interest in leveraging fifth generation (5G) wireless connectivity and mobile edge computing for offloading this processing to the cloud. To assess this possibility, this paper presents a detailed simulation and evaluation of 5G wireless offloading for object detection within a powerful, new smart wearable called VIS4ION, for the Blind-and-Visually Impaired (BVI). The current VIS4ION system is an instrumented book-bag with high-resolution cameras, vision processing and haptic and audio feedback. The paper considers uploading the camera data to a mobile edge cloud to perform real-time object detection and transmitting the detection results back to the wearable. To determine the video requirements, the paper evaluates the impact of video bit rate and resolution on object detection accuracy and range. A new street scene dataset with labeled objects relevant to BVI navigation is leveraged for analysis. The vision evaluation is combined with a detailed full-stack wireless network simulation to determine the distribution of throughputs and delays with real navigation paths and ray-tracing from new high-resolution 3D models in an urban environment. For comparison, the wireless simulation considers both a standard 4G-Long Term Evolution (LTE) carrier and high-rate 5G millimeter-wave (mmWave) carrier. The work thus provides a thorough and realistic assessment of edge computing with mmWave connectivity in an application with both high bandwidth and low latency requirements.
Contrastive language-image pretraining (CLIP) using image-text pairs has achieved impressive results on image classification in both zero-shot and transfer learning settings. However, we show that directly applying such models to recognize image regions for object detection leads to poor performance due to a domain shift: CLIP was trained to match an image as a whole to a text description, without capturing the fine-grained alignment between image regions and text spans. To mitigate this issue, we propose a new method called RegionCLIP that significantly extends CLIP to learn region-level visual representations, thus enabling fine-grained alignment between image regions and textual concepts. Our method leverages a CLIP model to match image regions with template captions and then pretrains our model to align these region-text pairs in the feature space. When transferring our pretrained model to the open-vocabulary object detection tasks, our method significantly outperforms the state of the art by 3.8 AP50 and 2.2 AP for novel categories on COCO and LVIS datasets, respectively. Moreoever, the learned region representations support zero-shot inference for object detection, showing promising results on both COCO and LVIS datasets. Our code is available at https://github.com/microsoft/RegionCLIP.
In recent years, GoogleNet has garnered substantial attention as one of the base convolutional neural networks (CNNs) to extract visual features for object detection. However, it experiences challenges of contaminated deep features when concatenating elements with different properties. Also, since GoogleNet is not an entirely lightweight CNN, it still has many execution overheads to apply to a resource-starved application domain. Therefore, a new CNNs, FactorNet, has been proposed to overcome these functional challenges. The FactorNet CNN is composed of multiple independent sub CNNs to encode different aspects of the deep visual features and has far fewer execution overheads in terms of weight parameters and floating-point operations. Incorporating FactorNet into the Faster-RCNN framework proved that FactorNet gives \ignore{a 5\%} better accuracy at a minimum and produces additional speedup over GoolgleNet throughout the KITTI object detection benchmark data set in a real-time object detection system.
Performance of object detection models has been growing rapidly on two major fronts, model accuracy and efficiency. However, in order to map deep neural network (DNN) based object detection models to edge devices, one typically needs to compress such models significantly, thus compromising the model accuracy. In this paper, we propose a novel edge GPU friendly module for multi-scale feature interaction by exploiting missing combinatorial connections between various feature scales in existing state-of-the-art methods. Additionally, we propose a novel transfer learning backbone adoption inspired by the changing translational information flow across various tasks, designed to complement our feature interaction module and together improve both accuracy as well as execution speed on various edge GPU devices available in the market. For instance, YOLO-ReT with MobileNetV2x0.75 backbone runs real-time on Jetson Nano, and achieves 68.75 mAP on Pascal VOC and 34.91 mAP on COCO, beating its peers by 3.05 mAP and 0.91 mAP respectively, while executing faster by 3.05 FPS. Furthermore, introducing our multi-scale feature interaction module in YOLOv4-tiny and YOLOv4-tiny (3l) improves their performance to 41.5 and 48.1 mAP respectively on COCO, outperforming the original versions by 1.3 and 0.9 mAP.
Depth is the hallmark of deep neural networks. But more depth means more sequential computation and higher latency. This begs the question -- is it possible to build high-performing "non-deep" neural networks? We show that it is. To do so, we use parallel subnetworks instead of stacking one layer after another. This helps effectively reduce depth while maintaining high performance. By utilizing parallel substructures, we show, for the first time, that a network with a depth of just 12 can achieve top-1 accuracy over 80% on ImageNet, 96% on CIFAR10, and 81% on CIFAR100. We also show that a network with a low-depth (12) backbone can achieve an AP of 48% on MS-COCO. We analyze the scaling rules for our design and show how to increase performance without changing the network's depth. Finally, we provide a proof of concept for how non-deep networks could be used to build low-latency recognition systems. Code is available at https://github.com/imankgoyal/NonDeepNetworks.
3D object detection using LiDAR data remains a key task for applications like autonomous driving and robotics. Unlike in the case of 2D images, LiDAR data is almost always collected over a period of time. However, most work in this area has focused on performing detection independent of the temporal domain. In this paper we present 3D-FCT, a Siamese network architecture that utilizes temporal information to simultaneously perform the related tasks of 3D object detection and tracking. The network is trained to predict the movement of an object based on the correlation features of extracted keypoints across time. Calculating correlation across keypoints only allows for real-time object detection. We further extend the multi-task objective to include a tracking regression loss. Finally, we produce high accuracy detections by linking short-term object tracklets into long term tracks based on the predicted tracks. Our proposed method is evaluated on the KITTI tracking dataset where it is shown to provide an improvement of 5.57% mAP over a state-of-the-art approach.
Deep Neural Networks (DNNs) are ubiquitous and span a variety of applications ranging from image classification to real-time object detection. As DNN models become more sophisticated, the computational cost of training these models becomes a burden. For this reason, outsourcing the training process has been the go-to option for many DNN users. Unfortunately, this comes at the cost of vulnerability to backdoor attacks. These attacks aim to establish hidden backdoors in the DNN so that it performs well on clean samples, but outputs a particular target label when a trigger is applied to the input. Existing backdoor attacks either generate triggers in the spatial domain or naively poison frequencies in the Fourier domain. In this work, we propose a pipeline based on Fourier heatmaps to generate a spatially dynamic and invisible backdoor attack in the frequency domain. The proposed attack is extensively evaluated on various datasets and network architectures. Unlike most existing backdoor attacks, the proposed attack can achieve high attack success rates with low poisoning rates and little to no drop in performance while remaining imperceptible to the human eye. Moreover, we show that the models poisoned by our attack are resistant to various state-of-the-art (SOTA) defenses, so we contribute two possible defenses that can evade the attack.
Efficient processing of high-res video streams is safety-critical for many robotics applications such as autonomous driving. To maintain real-time performance, many practical systems downsample the video stream. But this can hurt downstream tasks such as (small) object detection. Instead, we take inspiration from biological vision systems that allocate more foveal "pixels" to salient parts of the scene. We introduce FOVEA, an approach for intelligent downsampling that ensures salient image regions remain "magnified" in the downsampled output. Given a high-res image, FOVEA applies a differentiable resampling layer that outputs a small fixed-size image canvas, which is then processed with a differentiable vision module (e.g., object detection network), whose output is then differentiably backward mapped onto the original image size. The key idea is to resample such that background pixels can make room for salient pixels of interest. In order to ensure the overall pipeline remains efficient, FOVEA makes use of cheap and readily available cues for saliency, including dataset-specific spatial priors or temporal priors computed from object predictions in the recent past. On the autonomous driving datasets Argoverse-HD and BDD100K, our proposed method boosts the detection AP over standard Faster R-CNN, both with and without finetuning. Without any noticeable increase in compute, we improve accuracy on small objects by over 2x without degrading performance on large objects. Finally, FOVEA sets a new record for streaming AP (from 17.8 to 23.0 on a GTX 1080 Ti GPU), a metric designed to capture both accuracy and latency.
In this report, we introduce our real-time 2D object detection system for the realistic autonomous driving scenario. Our detector is built on a newly designed YOLO model, called YOLOX. On the Argoverse-HD dataset, our system achieves 41.0 streaming AP, which surpassed second place by 7.8/6.1 on detection-only track/fully track, respectively. Moreover, equipped with TensorRT, our model achieves the 30FPS inference speed with a high-resolution input size (e.g., 1440-2304). Code and models will be available at https://github.com/Megvii-BaseDetection/YOLOX
Deep Neural Network (DNN) trained object detectors are widely deployed in many mission-critical systems for real time video analytics at the edge, such as autonomous driving and video surveillance. A common performance requirement in these mission-critical edge services is the near real-time latency of online object detection on edge devices. However, even with well-trained DNN object detectors, the online detection quality at edge may deteriorate for a number of reasons, such as limited capacity to run DNN object detection models on heterogeneous edge devices, and detection quality degradation due to random frame dropping when the detection processing rate is significantly slower than the incoming video frame rate. This paper addresses these problems by exploiting multi-model multi-device detection parallelism for fast object detection in edge systems with heterogeneous edge devices. First, we analyze the performance bottleneck of running a well-trained DNN model at edge for real time online object detection. We use the offline detection as a reference model, and examine the root cause by analyzing the mismatch among the incoming video streaming rate, video processing rate for object detection, and output rate for real time detection visualization of video streaming. Second, we study performance optimizations by exploiting multi-model detection parallelism. We show that the model-parallel detection approach can effectively speed up the FPS detection processing rate, minimizing the FPS disparity with the incoming video frame rate on heterogeneous edge devices. We evaluate the proposed approach using SSD300 and YOLOv3 on benchmark videos of different video stream rates. The results show that exploiting multi-model detection parallelism can speed up the online object detection processing rate and deliver near real-time object detection performance for efficient video analytics at edge.
In this report, we present some experienced improvements to YOLO series, forming a new high-performance detector -- YOLOX. We switch the YOLO detector to an anchor-free manner and conduct other advanced detection techniques, i.e., a decoupled head and the leading label assignment strategy SimOTA to achieve state-of-the-art results across a large scale range of models: For YOLO-Nano with only 0.91M parameters and 1.08G FLOPs, we get 25.3% AP on COCO, surpassing NanoDet by 1.8% AP; for YOLOv3, one of the most widely used detectors in industry, we boost it to 47.3% AP on COCO, outperforming the current best practice by 3.0% AP; for YOLOX-L with roughly the same amount of parameters as YOLOv4-CSP, YOLOv5-L, we achieve 50.0% AP on COCO at a speed of 68.9 FPS on Tesla V100, exceeding YOLOv5-L by 1.8% AP. Further, we won the 1st Place on Streaming Perception Challenge (Workshop on Autonomous Driving at CVPR 2021) using a single YOLOX-L model. We hope this report can provide useful experience for developers and researchers in practical scenes, and we also provide deploy versions with ONNX, TensorRT, NCNN, and Openvino supported. Source code is at https://github.com/Megvii-BaseDetection/YOLOX.
This survey paper specially analyzed computer vision-based object detection challenges and solutions by different techniques. We mainly highlighted object detection by three different trending strategies, i.e., 1) domain adaptive deep learning-based approaches (discrepancy-based, Adversarial-based, Reconstruction-based, Hybrid). We examined general as well as tiny object detection-related challenges and offered solutions by historical and comparative analysis. In part 2) we mainly focused on tiny object detection techniques (multi-scale feature learning, Data augmentation, Training strategy (TS), Context-based detection, GAN-based detection). In part 3), To obtain knowledge-able findings, we discussed different object detection methods, i.e., convolutions and convolutional neural networks (CNN), pooling operations with trending types. Furthermore, we explained results with the help of some object detection algorithms, i.e., R-CNN, Fast R-CNN, Faster R-CNN, YOLO, and SSD, which are generally considered the base bone of CV, CNN, and OD. We performed comparative analysis on different datasets such as MS-COCO, PASCAL VOC07,12, and ImageNet to analyze results and present findings. At the end, we showed future directions with existing challenges of the field. In the future, OD methods and models can be analyzed for real-time object detection, tracking strategies.
Modern top-performing object detectors depend heavily on backbone networks, whose advances bring consistent performance gains through exploring more effective network structures. In this paper, we propose a novel and flexible backbone framework, namely CBNetV2, to construct high-performance detectors using existing open-sourced pre-trained backbones under the pre-training fine-tuning paradigm. In particular, CBNetV2 architecture groups multiple identical backbones, which are connected through composite connections. Specifically, it integrates the high- and low-level features of multiple backbone networks and gradually expands the receptive field to more efficiently perform object detection. We also propose a better training strategy with assistant supervision for CBNet-based detectors. Without additional pre-training of the composite backbone, CBNetV2 can be adapted to various backbones (CNN-based vs. Transformer-based) and head designs of most mainstream detectors (one-stage vs. two-stage, anchor-based vs. anchor-free-based). Experiments provide strong evidence that, compared with simply increasing the depth and width of the network, CBNetV2 introduces a more efficient, effective, and resource-friendly way to build high-performance backbone networks. Particularly, our Dual-Swin-L achieves 59.4% box AP and 51.6% mask AP on COCO test-dev under the single-model and single-scale testing protocol, which is significantly better than the state-of-the-art result (57.7% box AP and 50.2% mask AP) achieved by Swin-L, while the training schedule is reduced by 6$\times$. With multi-scale testing, we push the current best single model result to a new record of 60.1% box AP and 52.3% mask AP without using extra training data. Code is available at https://github.com/VDIGPKU/CBNetV2.
Object detection plays an important role in self-driving cars for security development. However, mobile systems on self-driving cars with limited computation resources lead to difficulties for object detection. To facilitate this, we propose a compiler-aware neural pruning search framework to achieve high-speed inference on autonomous vehicles for 2D and 3D object detection. The framework automatically searches the pruning scheme and rate for each layer to find a best-suited pruning for optimizing detection accuracy and speed performance under compiler optimization. Our experiments demonstrate that for the first time, the proposed method achieves (close-to) real-time, 55ms and 99ms inference times for YOLOv4 based 2D object detection and PointPillars based 3D detection, respectively, on an off-the-shelf mobile phone with minor (or no) accuracy loss.
When producing a model to object detection in a specific context, the first obstacle is to have a dataset labeling the desired classes. In RoboCup, some leagues already have more than one dataset to train and evaluate a model. However, in the Small Size League (SSL), there is not such dataset available yet. This paper presents an open-source dataset to be used as a benchmark for real-time object detection in SSL. This work also presented a pipeline to train, deploy, and evaluate Convolutional Neural Networks (CNNs) models in a low-power embedded system. This pipeline was used to evaluate the proposed dataset with state-of-art optimized models. In this dataset, the MobileNet SSD v1 achieves 44.88% AP (68.81% AP50) at 94 Frames Per Second (FPS) while running on an SSL robot.