The desire to empower resource-limited edge devices with computer vision (CV) must overcome the high energy consumption of collecting and processing vast sensory data. To address the challenge, this work proposes an energy-efficient non-von-Neumann in-pixel processing solution for neuromorphic vision sensors employing emerging (X) magnetic domain wall magnetic tunnel junction (MDWMTJ) for the first time, in conjunction with CMOS-based neuromorphic pixels. Our hybrid CMOS+X approach performs in-situ massively parallel asynchronous analog convolution, exhibiting low power consumption and high accuracy across various CV applications by leveraging the non-volatility and programmability of the MDWMTJ. Moreover, our developed device-circuit-algorithm co-design framework captures device constraints (low tunnel-magnetoresistance, low dynamic range) and circuit constraints (non-linearity, process variation, area consideration) based on monte-carlo simulations and device parameters utilizing GF22nm FD-SOI technology. Our experimental results suggest we can achieve an average of 45.3% reduction in backend-processor energy, maintaining similar front-end energy compared to the state-of-the-art and high accuracy of 79.17% and 95.99% on the DVS-CIFAR10 and IBM DVS128-Gesture datasets, respectively.
Transformer models have demonstrated high accuracy in numerous applications but have high complexity and lack sequential processing capability making them ill-suited for many streaming applications at the edge where devices are heavily resource-constrained. Thus motivated, many researchers have proposed reformulating the transformer models as RNN modules which modify the self-attention computation with explicit states. However, these approaches often incur significant performance degradation. The ultimate goal is to develop a model that has the following properties: parallel training, streaming and low-cost inference, and SOTA performance. In this paper, we propose a new direction to achieve this goal. We show how architectural modifications to a recurrent model can help push its performance toward Transformer models while retaining its sequential processing capability. Specifically, inspired by the recent success of Legendre Memory Units (LMU) in sequence learning tasks, we propose LMUFormer, which augments the LMU with convolutional patch embedding and convolutional channel mixer. Moreover, we present a spiking version of this architecture, which introduces the benefit of states within the patch embedding and channel mixer modules while simultaneously reducing the computing complexity. We evaluated our architectures on multiple sequence datasets. In comparison to SOTA transformer-based models within the ANN domain on the SCv2 dataset, our LMUFormer demonstrates comparable performance while necessitating a remarkable 53 times reduction in parameters and a substantial 65 times decrement in FLOPs. Additionally, owing to our model's proficiency in real-time data processing, we can achieve a 32.03% reduction in sequence length, all while incurring an inconsequential decline in performance. Our code is publicly available at https://github.com/zeyuliu1037/LMUFormer.git.
Bio-inspired Spiking Neural Networks (SNN) are now demonstrating comparable accuracy to intricate convolutional neural networks (CNN), all while delivering remarkable energy and latency efficiency when deployed on neuromorphic hardware. In particular, ANN-to-SNN conversion has recently gained significant traction in developing deep SNNs with close to state-of-the-art (SOTA) test accuracy on complex image recognition tasks. However, advanced ANN-to-SNN conversion approaches demonstrate that for lossless conversion, the number of SNN time steps must equal the number of quantization steps in the ANN activation function. Reducing the number of time steps significantly increases the conversion error. Moreover, the spiking activity of the SNN, which dominates the compute energy in neuromorphic chips, does not reduce proportionally with the number of time steps. To mitigate the accuracy concern, we propose a novel ANN-to-SNN conversion framework, that incurs an exponentially lower number of time steps compared to that required in the SOTA conversion approaches. Our framework modifies the SNN integrate-and-fire (IF) neuron model with identical complexity and shifts the bias term of each batch normalization (BN) layer in the trained ANN. To mitigate the spiking activity concern, we propose training the source ANN with a fine-grained L1 regularizer with surrogate gradients that encourages high spike sparsity in the converted SNN. Our proposed framework thus yields lossless SNNs with ultra-low latency, ultra-low compute energy, thanks to the ultra-low timesteps and high spike sparsity, and ultra-high test accuracy, for example, 73.30% with only 4 time steps on the ImageNet dataset.
Spiking Neural Networks (SNNs) have emerged as a popular spatio-temporal computing paradigm for complex vision tasks. Recently proposed SNN training algorithms have significantly reduced the number of time steps (down to 1) for improved latency and energy efficiency, however, they target only convolutional neural networks (CNN). These algorithms, when applied on the recently spotlighted vision transformers (ViT), either require a large number of time steps or fail to converge. Based on analysis of the histograms of the ANN and SNN activation maps, we hypothesize that each ViT block has a different sensitivity to the number of time steps. We propose a novel training framework that dynamically allocates the number of time steps to each ViT module depending on a trainable score assigned to each timestep. In particular, we generate a scalar binary time step mask that filters spikes emitted by each neuron in a leaky-integrate-and-fire (LIF) layer. The resulting SNNs have high activation sparsity and require only accumulate operations (AC), except for the input embedding layer, in contrast to expensive multiply-and-accumulates (MAC) needed in traditional ViTs. This yields significant improvements in energy efficiency. We evaluate our training framework and resulting SNNs on image recognition tasks including CIFAR10, CIFAR100, and ImageNet with different ViT architectures. We obtain a test accuracy of 95.97% with 4.97 time steps with direct encoding on CIFAR10.
Efficient and effective on-line detection and correction of bad pixels can improve yield and increase the expected lifetime of image sensors. This paper presents a comprehensive Deep Learning (DL) based on-line detection-correction approach, suitable for a wide range of pixel corruption rates. A confidence calibrated segmentation approach is introduced, which achieves nearly perfect bad pixel detection, even with few training samples. A computationally light-weight correction algorithm is proposed for low rates of pixel corruption, that surpasses the accuracy of traditional interpolation-based techniques. We also propose an autoencoder based image reconstruction approach which alleviates the need for prior bad pixel detection and yields promising results for high rates of pixel corruption. Unlike previous methods, which use proprietary images, we demonstrate the efficacy of the proposed methods on the open-source Samsung S7 ISP and MIT-Adobe FiveK datasets. Our approaches yield up to 99.6% detection accuracy with <0.6% false positives and corrected images within 1.5% average pixel error from 70% corrupted images.
Computer vision (CV) pipelines are typically evaluated on datasets processed by image signal processing (ISP) pipelines even though, for resource-constrained applications, an important research goal is to avoid as many ISP steps as possible. In particular, most CV datasets consist of global shutter (GS) images even though most cameras today use a rolling shutter (RS). This paper studies the impact of different shutter mechanisms on machine learning (ML) object detection models on a synthetic dataset that we generate using the advanced simulation capabilities of Unreal Engine 5 (UE5). In particular, we train and evaluate mainstream detection models with our synthetically-generated paired GS and RS datasets to ascertain whether there exists a significant difference in detection accuracy between these two shutter modalities, especially when capturing low-speed objects (e.g., pedestrians). The results of this emulation framework indicate the performance between them are remarkably congruent for coarse-grained detection (mean average precision (mAP) for IOU=0.5), but have significant differences for fine-grained measures of detection accuracy (mAP for IOU=0.5:0.95). This implies that ML pipelines might not need explicit correction for RS for many object detection applications, but mitigating RS effects in ISP-less ML pipelines that target fine-grained location of the objects may need additional research.
This paper presents "FireFly", a synthetic dataset for ember detection created using Unreal Engine 4 (UE4), designed to overcome the current lack of ember-specific training resources. To create the dataset, we present a tool that allows the automated generation of the synthetic labeled dataset with adjustable parameters, enabling data diversity from various environmental conditions, making the dataset both diverse and customizable based on user requirements. We generated a total of 19,273 frames that have been used to evaluate FireFly on four popular object detection models. Further to minimize human intervention, we leveraged a trained model to create a semi-automatic labeling process for real-life ember frames. Moreover, we demonstrated an up to 8.57% improvement in mean Average Precision (mAP) in real-world wildfire scenarios compared to models trained exclusively on a small real dataset.
The massive amounts of data generated by camera sensors motivate data processing inside pixel arrays, i.e., at the extreme-edge. Several critical developments have fueled recent interest in the processing-in-pixel-in-memory paradigm for a wide range of visual machine intelligence tasks, including (1) advances in 3D integration technology to enable complex processing inside each pixel in a 3D integrated manner while maintaining pixel density, (2) analog processing circuit techniques for massively parallel low-energy in-pixel computations, and (3) algorithmic techniques to mitigate non-idealities associated with analog processing through hardware-aware training schemes. This article presents a comprehensive technology-circuit-algorithm landscape that connects technology capabilities, circuit design strategies, and algorithmic optimizations to power, performance, area, bandwidth reduction, and application-level accuracy metrics. We present our results using a comprehensive co-design framework incorporating hardware and algorithmic optimizations for various complex real-life visual intelligence tasks mapped onto our P2M paradigm.
Vision Transformer models, such as ViT, Swin Transformer, and Transformer-in-Transformer, have recently gained significant traction in computer vision tasks due to their ability to capture the global relation between features which leads to superior performance. However, they are compute-heavy and difficult to deploy in resource-constrained edge devices. Existing hardware accelerators, including those for the closely-related BERT transformer models, do not target highly resource-constrained environments. In this paper, we address this gap and propose ViTA - a configurable hardware accelerator for inference of vision transformer models, targeting resource-constrained edge computing devices and avoiding repeated off-chip memory accesses. We employ a head-level pipeline and inter-layer MLP optimizations, and can support several commonly used vision transformer models with changes solely in our control logic. We achieve nearly 90% hardware utilization efficiency on most vision transformer models, report a power of 0.88W when synthesised with a clock of 150 MHz, and get reasonable frame rates - all of which makes ViTA suitable for edge applications.
Edge devices equipped with computer vision must deal with vast amounts of sensory data with limited computing resources. Hence, researchers have been exploring different energy-efficient solutions such as near-sensor processing, in-sensor processing, and in-pixel processing, bringing the computation closer to the sensor. In particular, in-pixel processing embeds the computation capabilities inside the pixel array and achieves high energy efficiency by generating low-level features instead of the raw data stream from CMOS image sensors. Many different in-pixel processing techniques and approaches have been demonstrated on conventional frame-based CMOS imagers, however, the processing-in-pixel approach for neuromorphic vision sensors has not been explored so far. In this work, we for the first time, propose an asynchronous non-von-Neumann analog processing-in-pixel paradigm to perform convolution operations by integrating in-situ multi-bit multi-channel convolution inside the pixel array performing analog multiply and accumulate (MAC) operations that consume significantly less energy than their digital MAC alternative. To make this approach viable, we incorporate the circuit's non-ideality, leakage, and process variations into a novel hardware-algorithm co-design framework that leverages extensive HSpice simulations of our proposed circuit using the GF22nm FD-SOI technology node. We verified our framework on state-of-the-art neuromorphic vision sensor datasets and show that our solution consumes ~2x lower backend-processor energy while maintaining almost similar front-end (sensor) energy on the IBM DVS128-Gesture dataset than the state-of-the-art while maintaining a high test accuracy of 88.36%.