Perception Understanding Action: Adding Understanding to the Perception Action Cycle With Spiking Segmentation

Kirkland, Paul; Caterina, Gaetano Di; Soraghan, John J.; Matich, George

doi:10.3389/fnbot.2020.568319

Cited by 12 publications

(19 citation statements)

References 52 publications

(58 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This was to limit the number of external factors and focus the testing on the intra-classification abilities rather than inter classification. To further validate the SMASH method the 5 and 10 class networks where also utilised from 95 100 100 [35]. These tests helps to validate the SMASH method in a more complex feature similarity environment, where each class has less representation in layer Conv 2, as these networks utilised 16 features times the number of classes, 80 and 160 for the 5 and 10 classes respectively.…”

Section: Resultsmentioning

confidence: 99%

“…The proposed temporal spike matching algorithm HULK SMASH makes use of an image classification SNN trained using STDP, similar to that seen in [31,32,33]. Prior developments have shown that extending an image classification SNN to a semantic segmentation network was possible [34,35]. These two prior works also utilised the N-Caltech dataset [36] making use of the event driven nature of the input sequences.…”

Section: Temporal Spike Matchingmentioning

confidence: 99%

“…The featural data then has the same total number of features as the network, which in this case, is 41. This value comes from the use of a network with a similar structure as the Face-Motorbike Network used Kirkland et al [35], except with only one classification layer. The other 40 features come from the 36 from Trans-Conv2 and 4 from Trans-Conv1.…”

Section: Similarity Matching Through Active Spike Hashingmentioning

confidence: 99%

“…This allows the overall interpretability of the network to increase, with the decisions made by the network to become viewable and assessable, so removing the black box stigma of neural networks. This work makes use of previous research that showed how an SCNN trained with STDP could first perform image classification [33], and recently how it can be transformed into an encoder-decoder network, allowing semantic segmentation to be realised [34,35]. However, this previous research could not distinguish instances within these semantic segmentations.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Unsupervised Spiking Instance Segmentation on Event Data using STDP

Kirkland¹,

Manna²,

Vicente-Sola³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Spiking Neural Networks (SNN) and the field of Neuromorphic Engineering has brought about a paradigm shift in how to approach Machine Learning (ML) and Computer Vision (CV) problem. This paradigm shift comes from the adaption of event-based sensing and processing. An event-based vision sensor allows for sparse and asynchronous events to be produced that are dynamically related to the scene. Allowing not only the spatial information but a high-fidelity of temporal information to be captured. Meanwhile avoiding the extra overhead and redundancy of conventional high frame rate approaches. However, with this change in paradigm, many techniques from traditional CV and ML are not applicable to these event-based spatial-temporal visual streams. As such a limited number of recognition, detection and segmentation approaches exist. In this paper, we present a novel approach that can perform instance segmentation using just the weights of a Spike Time Dependent Plasticity trained Spiking Convolutional Neural Network that was trained for object recognition. This exploits the spatial and temporal aspects of the network's internal feature representations adding this new discriminative capability. We highlight the new capability by successfully transforming a single class unsupervised network for face detection into a multi-person face recognition and instance segmentation network.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Temporal Spike Matchingmentioning

confidence: 99%

Section: Similarity Matching Through Active Spike Hashingmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Unsupervised Spiking Instance Segmentation on Event Data using STDP

Kirkland¹,

Manna²,

Vicente-Sola³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…The last case study highlights the results from the use of the previous two, together with an action based tracking and control system, 54 where the previous segmentation methodology is now built into an end-to-end NM system for following a chosen class within a scene. This describes how to use the results of the semantic segmentation to help to provide a control output for a system, such a camera on a pan tilt mount, that could be used to tracking a class as it moves around a scene.…”

Section: Tracking and Controlmentioning

confidence: 99%

Neuromorphic technologies for defence and security

Kirkland

Caterina

Soraghan

et al. 2020

Emerging Imaging and Sensing Technologies for Security and Defence V; And Advanced Manufacturing Technologies for Micro- And Na

Self Cite

View full text Add to dashboard Cite

Despite the highly promising advances in Machine Learning (ML) and Deep Learning (DL) in recent years, DL requires significant hardware acceleration to be effective, as it is rather computationally expensive. Moreover, miniaturisation of electronic devices requires small form-factor processing units, with reduced SWaP (Size, Weight and Power) profile. Therefore, a completely new processing paradigm is needed to address both issues. In this context, the concept of neuromorphic (NM) engineering provides an attractive alternative, seen as the analog/digital implementation of biologically brain inspired neural networks. NM systems propagate spikes as means of processing data, with the information being encoded in the timing and rate of spikes generated by each neuron of a so-called spiking neural network (SNN). Based on this, the key advantages of SNNs are: less computational power required, more efficient and faster processing, much lower power consumption. This paper reports on the current state of the art in the field of NM systems, and it describes three application scenarios of SNN-based processing for security and defence, namely target detection and tracking, semantic segmentation, and control.

show abstract