Optomotor response/reflex (OMR) assays are emerging as a powerful and versatile tool for phenotypic study and new drug discovery for eye and brain disorders. Yet efficient OMR assessment for visual performance in mice remains a challenge. Existing OMR testing devices for mice require a lengthy procedure and may be subject to bias due to use of artificial criteria. We developed an optimized staircase protocol that utilizes mouse head pausing behavior as a novel indicator for the absence of OMR, to allow rapid and unambiguous vision assessment. It provided a highly sensitive and reliable method that can be easily implemented into automated or manual OMR systems to allow quick and unbiased assessment for visual acuity and contrast sensitivity in mice. The sensitivity and quantitative capacity of the protocol were validated using wild type mice and an inherited mouse model of retinal degeneration – mice carrying rhodopsin deficiency and exhibiting progressive loss of photoreceptors. Our OMR system with this protocol was capable of detecting progressive visual function decline that was closely correlated with the loss of photoreceptors in rhodopsin deficient mice. It provides significant advances over the existing methods in the currently available OMR devices in terms of sensitivity, accuracy and efficiency.

This paper proposes a bio-inspired visual motion estimation algorithm based on motion energy, along with its compact very-large-scale integration (VLSI) architecture using low-cost embedded systems. The algorithm mimics motion perception functions of retina, V1, and MT neurons in a primate visual system. It involves operations of ternary edge extraction, spatiotemporal filtering, motion energy extraction, and velocity integration. Moreover, we propose the concept of confidence map to indicate the reliability of estimation results on each probing location. Our algorithm involves only additions and multiplications during runtime, which is suitable for low-cost hardware implementation. The proposed VLSI architecture employs multiple (frame, pixel, and operation) levels of pipeline and massively parallel processing arrays to boost the system performance. The array unit circuits are optimized to minimize hardware resource consumption. We have prototyped the proposed architecture on a low-cost field-programmable gate array platform (Zynq 7020) running at 53-MHz clock frequency. It achieved 30-frame/s real-time performance for velocity estimation on 160 × 120 probing locations. A comprehensive evaluation experiment showed that the estimated velocity by our prototype has relatively small errors (average endpoint error < 0.5 pixel and angular error < 10°) for most motion cases.

This paper presents a lightweight statistical learning framework potentially suitable for low-cost event-based vision systems, where visual information is captured by a dynamic vision sensor (DVS) and represented as an asynchronous stream of pixel addresses (events) indicating a relative intensity change on those locations. A simple random ferns classifier based on randomly selected patch-based binary features is employed to categorize pixel event flows. Our experimental results demonstrate that compared to existing event-based processing algorithms, such as spiking convolutional neural networks (SCNNs) and the state-of-the-art bag-of-events (BoE)-based statistical algorithms, our framework excels in high processing speed (2× faster than the BoE statistical methods and >100× faster than previous SCNNs in training speed) with extremely simple online learning process, and achieves state-of-the-art classification accuracy on four popular address-event representation data sets: MNIST-DVS, Poker-DVS, Posture-DVS, and CIFAR10-DVS. Hardware estimation shows that our algorithm will be preferable for low-cost embedded system implementations.

No abstract

No abstract

No abstract

This work proposes a hardware-friendly, dense optical flow-based Time-to-Collision (TTC) estimation algorithm intended to be deployed on smart video sensors for collision avoidance. The algorithm optimized for hardware first extracts biological visual motion features (motion energies), and then utilizes a Random Forests regressor to predict robust and dense optical flow. Finally, TTC is reliably estimated from the divergence of the optical flow field. This algorithm involves only feed-forward data flows with simple pixel-level operations, and hence has inherent parallelism for hardware acceleration. The algorithm offers good scalability, allowing for flexible tradeoffs among estimation accuracy, processing speed and hardware resource. Experimental evaluation shows that the accuracy of the optical flow estimation is improved due to the use of Random Forests compared to existing voting-based approaches. Furthermore, results show that estimated TTC values by the algorithm closely follow the ground truth. The specifics of the hardware design to implement the algorithm on a real-time embedded system are laid out.

Spiking neural networks (SNNs) have attracted intensive attention due to the efficient event-driven computing paradigm. Among SNN training methods, the ANN-to-SNN conversion is usually regarded to achieve state-of-the-art recognition accuracies. However, many existing ANN-to-SNN techniques impose lengthy post-conversion steps like threshold balancing and weight renormalization, to compensate for the inherent behavioral discrepancy between artificial and spiking neurons. In addition, they require a long temporal window to encode and process as many spikes as possible to better approximate the real-valued ANN neurons, leading to a high inference latency. To overcome these challenges, we propose a calcium-gated bipolar leaky integrate and fire (Ca-LIF) spiking neuron model to better approximate the functions of the ReLU neurons widely adopted in ANNs. We also propose a quantization-aware training (QAT)-based framework leveraging an off-the-shelf QAT toolkit for easy ANN-to-SNN conversion, which directly exports the learned ANN weights to SNNs requiring no post-conversion processing. We benchmarked our method on typical deep network structures with varying time-step lengths from 8 to 128. Compared to other research, our converted SNNs reported competitively high-accuracy performance, while enjoying relatively short inference time steps.