Westfälische Wilhelms-Universität Muenster (WWU) PhD researcher Jiaqi Zhang, recently completed a 6 month secondment for ULTRACEPT at partner Institute of Automation, Chinese Academy of Sciences (CASIA) in China from October 2021 to May 2022. Jiaqi shares her secondment experience in this blog post.

During my 6 month secondment at the Institute of Automation Chinese Academy of Sciences (CASIA), I focused on my research on interpolation kernel machine, which is a fundamental machine learning method that can be used for classification and regression problems.

Interpolation kernel machine is a general class of kernel-based techniques for classification and regression with several favorable properties. In its standard formulation, it can only deal with positive definite kernels. In real-world applications, however, one often has good reasons or even has to work with indefinite kernels.

While many kernel based machine learning methods, in particular support vector machines, have seen extensions to indefinite kernels, there was, to our knowledge, no previous work on indefinite interpolation kernel machines yet. We address this open issue and propose strategies and concrete implementations to make interpolation kernel machines functional for indefinite kernels, which is our main contribution. We conducted experiments on graph datasets from various domains and UCI datasets show competitive performance, also in comparison with the indefinite SVM counterparts. Currently, interpolation kernel machines have not received due attention, in particular, compared to SVM. Our work thus will help to increase the awareness of this powerful learning model in the community as well.

I am very appreciative of the ULTRACEPT project, Prof. Liu (CASIA), Prof. Jiang (WWU), and the EU H2020 funding which provided me the opportunity to work with CASIA. The Pattern Analysis and Learning group (PAL) has skilled and kind researchers, as well as significant research contributions.

During my 6 month secondment, I improved my skills in finding and solving problems. I learned about the research environment in China, which helped me to choose my future career directions. It also enabled me to understand the difference in research strategies between Germany and China, which can help me to enrich my own work skills and make decisions for my future career.

You can read more about Jiaqi’s research in her publication which she presented at the ICPRAI 2022 – 3rd International Conference on Pattern Recognition and Artificial Intelligence held June 1- 3, 2022 in Paris, France. She presented during the Special Session: Graphs for Pattern Recognition: Representations, Theory and Applications.

Zhang, J., Liu, CL., Jiang, X. (2022). Interpolation Kernel Machine and Indefinite Kernel Methods for Graph Classification. In: El Yacoubi, M., Granger, E., Yuen, P.C., Pal, U., Vincent, N. (eds) Pattern Recognition and Artificial Intelligence. ICPRAI 2022. Lecture Notes in Computer Science, vol 13364. Springer, Cham. https://doi.org/10.1007/978-3-031-09282-4_39

University of Lincoln Masters researcher Mu Hua, recently completed a 12 month secondment for ULTRACEPT at partner Guangzhou University in China.

During my one-year secondment at Guangzhou University, my research was based on the previous works on the locusts LGMD (Lobula Giant Movement Detector) neural networks for collision perception, including the LGMD1 from Prof Shigang Yue and LGMD2 from Dr Qinbing Fu. My work was mainly focused on improving the LGMDs neural network’s ability for ultra-fast approaching objects.

Benefiting from thousands of decades’ evolution, locusts have been equipped with a vision system that improves their success rate of evading their natural predators in the blink of an eye. Taking inspiration from nature through the computational models of LGMDs in locust’s visual pathways has had a positive impact on addressing these problems. However, it is still challenging for current LGMD neural networks to accurately and reliably recognize the imminent collision when the approaching object is ultra-fast (see Fig. 1). The green dashed line is the threshold we set to indicate whether collision is happening or not; the Blue curve is the current LGMD1 responses to the ultra-fast objects. The neuron fires spikes and generates a ‘false alert’ while the approaching black ball is far away.

Since the refractoriness, namely the refractory period which is a common mechanism within plenty of creatures’ neuron systems, is able to assist together with other sorts of mechanisms to help stabilize a neuron. It is then introduced to previous works of LGMDs neural networks for further improvement. On the left in Figure 2, we show a comparison between our new proposed LGMD1 neural network and the previous one from Shigang Yue. On the right, we demonstrate the comparison between our proposed method of LGMD2 and the previous one from Qinbing Fu.

To better understand the refractoriness mechanism and explain the rationality of integrating the LGMDs neural network, we sought guidance from professors, Prof. Jigen Peng and Prof. Huang, and our outstanding colleagues. Their inference from the perspective of mathematics supported the proposed method (see Fig.3).

During my secondment, I obtained knowledge on both bio-plausible neural networks and coding and gained much experience in setting up experiments and analysing the experimental results. Many thanks to the ULTRACEPT project for supporting my research at Guangzhou University, and to my host, Prof. Jigen Peng for kindly providing me access to his well-equipped lab.

Fang Lei enrolled as a PhD Scholar at the University of Lincoln in 2019. In early 2020 she visited Guangzhou University as part of the STEP2DYNA project funded by the European Union’s Horizon 2020 Research and Innovation Programme under the Marie Skolodowska-Curie grant agreement. During this secondment Fang Lei was working on developing bio-inspired visual systems for collision detection in dim light environments. More recently, Fang continued this work during her 12 month secondment at Guangzhou University under ULTRACEPT from May 2020 to 2021.

During the secondment to Guangzhou University, I was working on developing bio-inspired visual systems for collision detection in dim light environments. For the autonomous navigation of vehicles or robots, it is a challenging task to detect moving objects in extremely low-light conditions due to very low signal-to-noise ratios (SNRs). However, nocturnal insects possess remarkable visual abilities in perceiving motion cues and detecting moving objects in very dim light environments. There are many studies on the night vision of insects’ visual systems, which provide us with a lot of inspirations for enhancing motion cues and modelling an artificial visual system to detect motion like looming objects. Fig. 1 shows an example image of looming motion in a dim light environment which is from the low-light video motion (LLVM) dataset obtained by the experimental devices (see Fig. 2).

In order to develop more ideas and experiences in my modelling work, I discussed this with other colleagues and Prof. Peng (see Fig. 3) and got very useful suggestions. We discussed mainly the biological modelling of direction selectivity of LGMD1. We also organized a group seminar every week to discuss the related problems we encounter in our research projects, and I also got a lot of valuable experiences on bio-inspired modelling by sharing our ideas.

For my research work, collision detection in a dim light environment includes the modelling work of direction selectivity of LGMD1 neuron and the motion cues enhancement. I have developed the new LGMD1 model which is effective in distinguishing looming motion from translating motion. I have published one conference paper and attended the online virtual conference (IJCNN 2021, see Fig. 4). I also submitted one journal paper to IEEE transactions on neural networks and learning systems (NNLS) which is under review. Additionally, I have finished the modelling work of motion cues enhancement and proposed a new model. Fig. 5 shows the enhancement results of the captured dark image sequences during testing experiments.

During this 12-month secondment, I have a better knowledge of bio-inspired modelling and obtain a lot of exercises of connection between theory and practice.  I established good friendships with my colleagues through frequent communications in every week’s group seminar, which provide a basis for future cooperation. The secondment was a very precious experience for me. Many thanks to ULTRACEPT project for supporting my research work and providing me with the opportunity to work together with my colleagues.

Hongxin Wang received his PhD in computer science from the University of Lincoln in 2020. Following a secondment under the STEP2DYNA project, Dr Wang carried out a further secondment under the ULTRACEPT project from April 2020 to April 2021 at partner Guangzhou University. Here, he undertook research contributing to work packages 1 and 2. Dr Wang’s ULTRACEPT contributions have involved directing the research into computational modelling of motion vision neural systems for small target motion detection. 

University of Lincoln’s Experienced Researcher Dr Hongxin Wang recently completed a 12 month secondment at ULTRACEPT project partner Guangzhou University in China. The project is funded by the European Union’s Horizon 2020 Research and Innovation Program under the Marie Skolodowska-Curie grant agreement. Dr Wang visited Guangzhou from April 2020 to April 2021 and contributed to Work Package 1 and 2.

Dr Wang reflects on what he has achieved during secondment

Monitoring moving objects against complex natural backgrounds is a huge challenge to future robotic vision systems, and even more so when attempting to detect small targets only a few pixels in size, for example, an unmanned aerial vehicle (UAV) or a bird in the distance, as shown in Fig. 1. Surprisingly, insects are quite apt at searching for mates and tracking prey, which appears as small dim speckles in the visual field. The exquisite sensitivity of insects for small target motion comes from a class of specific neurons called small target motion detectors (STMDs). Building a quantitative STMD model is the first step for not only further understanding the biological visual system but also providing robust and economic solutions of small target detection for an artificial vision system.

During this twelve-month secondment, I continued my previous work on modeling insects’ visual systems for small target detection and have made great progress. Specifically, we proposed a STMD-based model with time-delay feedback to achieve superior detection performance for fast-moving small targets, whilst significantly suppressing background false positive movements which display lower velocities. This work has been submitted to IEEE Transactions on Neural Networks and Learning Systems and is currently under review. In addition, we developed an attention-prediction guided visual system to overcome the heavy dependency of the existing models on target contrast to background, as illustrated in Fig. 2. The paper presenting this work has been completed and will be submitted to IEEE Transactions on Cybernetics.

During my 12 month secondment at Guangzhou University, I obtained inspiration and mathematical theory support from Professor Jigen Peng to design the STMD-based visual systems. We organized a seminar every week to discuss the latest biological findings, explore effective neural modeling methods, and develop specialised mathematical theory for bioinspired motion detection. Significant progress was made under the help of Professor Jigen Peng.

The secondment has also provided me with an opportunity to improve my mathematical ability with support from Professor Peng. Strong mathematical ability helps me better describe the insects’ visual systems, and build robust neural models for small target motion detection. In addition, I established a deep friendship with Professor Peng and my colleagues at Guangzhou University, which is providing me a basis for future research collaborations. Lastly, I introduced our research to colleagues during the discussion, which may attract their attention to our research field and finally boost the development of neural system modelling.

The secondment has been an excellent experience for me and provided me the opportunity to collaborate with my project colleagues. Thank you for the support from the ULTRACEPT project which benefited me a lot.

Azreen Azman is an associate professor at the Universiti Putra Malaysia in Kuala Lumpur.  He has just completed a 6 month secondment at the University of Lincoln and a 6 month secondment at Visomorphic Technology Ltd as part of the ULTRACEPT project funded by the European Union’s Horizon 2020 Research and Innovation Programme under the Marie Skolodowska-Curie grant agreement. He has been involved in Work Packages 2 and 3.

Hazard perception and collision detection are important components for the safety of an autonomous car, and it becomes more challenging in low light environment. During the twelve month secondment period my focus was to investigate the method for the detection of objects on the road in low light conditions by using captured images or video in order to recognise hazards or avoid collision.

One of the first tasks Azreen conducted in Lincoln was to collect audio-visual data in different road conditions. Azreen had the opportunity to join his colleagues Siavash Bahrami and Assoc Prof Shyamala Doraisamy from UPM who were also carrying out ULTRACEPT secondments at UoL and conducting audio-visual recordings of the road at the Millbrook Proving Ground in Bedford, United Kingdom. This provided a controlled environment in addition to other recordings conducted on normal roads.

It is anticipated that the performance of deep-learning based object detection algorithms such as R-CNN variants and YoLo diminishes as the input images become darker, due to the reduced amount of light and increased noise in the captured images. In Azreen’s preliminary experiment which used the Faster R-CNN model trained and tested on a collection of self-collected road images, the object detection performance is significantly reduced to almost 81% for dark and noisy images, as compared to the daylight images.

To overcome the problem, an image enhancement and noise reduction method was applied to the dark images prior to the object detection module. In his investigations, Azreen trained the LLNet, a deep autoencoder based image enhancement and noise reduction method for dark image enhancement.  As a result, the Faster R-CNN is able to detect 29% more objects on the enhanced images as compared to the dark images. The performance of the deep learning-based LLNet is better than the conventional Histogram Equalisation (HE) and Retinex methods. However, the patches prediction and image reconstruction steps are computationally expensive for real-time applications.

In August 2020, Azreen began his secondment at Visomorphic Technology Ltd, an industry partner for the ULTRACEPT project. In collaboration with the team, he continued working on the model to improve its efficiency for real-time application. His focus was to adopt the principles of the nocturnal insect vision system for image enhancement and object detection.

During Azreen’s stay in the UK, he attended and presented at the annual ULTRACEPT mid-term project meeting which was held in February 2020 and hosted in Cambridge. Azreen presented his work ‘Detection of objects on the road in low light condition using deep learning’. He also participated in ULTRACEPT Sandpit Session 1 facilitated by Qinbing Fu.

In addition, Azreen attended the first Lincoln Conference on Intelligent Robots and Systems organised by Lincoln Centre of Autonomous Systems (L-CAS) and the Keynote Session delivered by Prof. Graham Kendall from the University of Nottingham on Hyper-heuristics, both held in October 2020.

‘The secondment has given me the opportunities and resources to conduct my research for the project and to improve my skills and networking though various meetings and discussions. Despite the challenges faced due to the ongoing pandemic, both of my hosts (University of Lincoln and Visomorphic Technology Ltd) have provided me with the support to work remotely while continuously engaging with other researchers virtually. I would like to thank the sponsors including Universiti Putra Malaysia  and the ULTRACEPT’s Marie Sklodowska-Curie secondment grant for these opportunities.’ Azreen Azman

Dr Qinbing Fu received his PhD at University of Lincoln, in October 2018. Following a secondment under the STEP2DYNA project, Dr Fu carried out a further secondment under the ULTRACEPT project from August 2019 to August 2020 at partner Guangzhou University. Here he undertook research contributing to work packages 1 and 4. Dr Fu then went on to work as a postdoctoral researcher with Professor Shigang Yue until January 2021. Dr Fu’s ULTRACEPT contributions have involved directing the research into computational modelling of motion vision neural systems and applications on robotics. His research achievements and outputs for this project thus far is outlined in this blog post.

Research Outcomes

In support of the ULTRACEPT project, Dr Fu has published seven research papers including five journal papers and two conference papers. He was the first author on five of the publications and co-authored the other two. His main achievements have included:

• The modelling of LGMD-1 and LGMD-2 collision perception neural network models with applications on robot and vehicle scenarios;
• The modelling of Drosophila motion vision neural system for decoding the direction of foreground translating object in moving cluttered background;
• A review on the related field of research;
• Multiple neural system models integration for collision sensing.

Dr Fu’s research outputs can be found on his personal web pages on Google Scholar and ResearchGate. In addition, Qinbing directed promising research on building visually dynamic walls in an arena to test the on-board visual system. These research ideas have been collated and summarised in his research papers.

Dr Fu’s research contributions have fully supported ULTRACEPT’s WP1 and WP4. This includes modelling work on collision detection visual systems with systematic experiments on vehicle scenarios and also the integration of multiple neural system models for motion perception.

Secondment at Guangzhou University, China

Dr Fu carried out his ULTRACEPT secondment at project partner GZHU in China where he worked with Professor Jigen Peng. During this period he developed his capability on several aspects, becoming a more mature researcher in the academic community. This included: aspiring to progressive research ideas, collaboration with group members on completing research papers, coordinating teamwork, disseminating the project, good communication experience with global partners, and writing project proposals. Undoubtedly, the ULTRACEPT secondment for Dr Fu has been very successful.

Dissemination Activities

Dr Fu has undertaken a number of dissemination activities to promote the ULTRACEPT research outcomes. On the 28th July 2020, he presented his research at the ULTRACEPT online Workshop 2 on the topic of “Adaptive Inhibition Matters to Robust Collision Perception in Highly Variable Environments”. At this event, he exchanged ideas with project partners.

Dr Fu also facilitated a ULTRACEPT online Sandpit Session on 27 November 2020, during where he gave a talk on “Past, Present, and Future Modelling on Bio-Inspired Collision Detection Visual Systems: Towards the Robust Perception”.

On 18th December 2020, Dr Fu attended the 2020 IEEE International Conference on Advanced Robotics and Mechatronics (ICARM) held in Shenzhen, China where he presented his research paper entitled “Complementary visual neuronal systems model for collision sensing”. He also chaired a Session on “Biomimetics” during this conference.

Tian Liu enrolled as a PhD Scholar at the University of Lincoln in 2018. In 2018-2019 he visited Guangzhou University as part of the STEP2DYNA project funded by the European Union’s Horizon 2020 Research and Innovation Programme under the Marie Skolodowska-Curie grant agreement. During this secondment Tian Liu developed the ColCOSΦ experiment platform for social insects and swarm robotic researching. Tian investigated how multiple virtual pheromones impact on the swarm robots. More recently, Tian completed a 12 month secondment under ULTRACEPT at Guangzhou University.

Tian Liu recently completed his second 12 month secondment at project partner Guangzhou University in China as part of the ULTRACEPT project funded by the European Union’s Horizon 2020 Research and Innovation Programme under the Marie Skolodowska-Curie grant agreement. Tian visited Guangzhou from November 2019 to November 2020 and has been involved in Work Package 1 and 4.

Tian reflects on what he has achieved during his time in Guangzhou

Most social insects, such as ants, only have a tiny brain. However, they can complete very difficult and complex tasks with a large number of individuals cooperating. Examples include building a large nest or collecting food through rugged routes. They are able to do this because the pheromones act as an important communication medium.

During this 12 month secondment, I continued to focus my attention on swarm robots with multiple pheromones. I believed that it is the interaction of multiple pheromones that enables insects to perform such demanding tasks, rather than the single pheromone mechanism which is now so widely studied. I worked with ULTRACEPT researcher Xuelong Sun and Dr. Cheng Hu to develop the ColCOSΦ, which can easily implement multiple pheromone research experiments. We verified the application and evaluation of the effects of multi-pheromones in swarm robotics by implementing several case studies which simulated ants foraging and carrying out hunting and deployment tasks.

I showcased the outcomes of this research at both ICARM2019 and ICARM2020 international conferences.

Due to its excellent scalability, we also use it for research experiments in related fields. For example, the platform can simulate traffic scenarios so we can test our LGMD model (a collision detection model) by using the micro robot (Colias) in a low-cost way.

Besides olfactory, the visual information is also a very important input for insects, so we implemented a changeable visual environment on the ColCOSΦ for investigating how to make full use of both olfactory and visual information in a swarm task. The research was collated into two articles which have been submitted to ICRA2021 with fellow ULTRACEPT researchers Xuelong Sun, Dr Qinbing Fu and Dr Cheng Hu.

The secondment has been an excellent experience for me and my colleagues and provided me the opportunity to collaborate with my project colleagues.

Many thanks to ULTRACEPT project for supporting my research and for allowing me to work with these outstanding research scholars.

Xuelong Sun enrolled as a PhD Scholar at the University of Lincoln in 2016. In 2017-18 he visited Tsinghua University, China as part of the STEP2DYNA project funded by the European Union’s Horizon 2020 Research and Innovation Program under the Marie Skolodowska-Curie grant agreement. During the secondment, Xuelong revisited the classical ring attractor model and demonstrated its application of bio-plausible optimal cue integration of directional cues. More recently, Xuelong completed a 12 month secondment with Guangzhou University under the ULTRACEPT project.

Xuelong Sun recently completed a 12 month secondment at project partner Guangzhou University in China as part of the ULTRACEPT project funded by the European Union’s Horizon 2020 Research and Innovation Program under the Marie Skolodowska-Curie grant agreement. Xuelongvisited Guangzhou from January 2019 to March 2019, then again from July 2019 to May 2020. Xuelong has been involved in Work Package 1 and 3.

Xuelong reflects on what he has achieved during his time in Guangzhou

Solving problems by taking inspiration from animals (so-called bio-inspired solutions) is one of the core ideas of our group-computational intelligence lab (CIL). As for me, insects are my best friend because of their amazing ability to use navigation and efficient collaboration to solve complex problems.

During this ten-month secondment, I continued my previous modeling work of insect navigation systems and have made great progress, by not only reproducing the main observed behavioural data of real insects, but also mapping specific computation to corresponding brain regions of the insects. We are making great contributions to the insect navigation community.

The paper presenting this work has been submited to eLife during the secondment (Dec 2019) and then has been accepted and published- A decentralised neural model explaining optimal integration of navigational strategies in insects. I also attended the online conference Neuronmatch Conference in March 2020 to present this work.

As part of my researching interests cooperating with fellow ULTRACEPT researcher Tian Liu, we developed a platform called ColCOSՓ for social insects and swarm robotic researching. This platform consists of three parts, the arena (LED screen), the monitoring camera, and the Colias micro-robot. Swarm robotic and social insects related experimental scenarios can be easily and flexibly conducted in this platform. Fellow ULTRACEPT researcher Dr Cheng Hu and I presented the platform physically at Guangdong (Foshan) Doctoral and Postdoctoral Talent Exchange and Technology Project Matchmaking Conference.

Another interesting experiment undertaken during my secondment is that we investigated the performance of LGMD model of collision avoidance in the context of city traffic. The real-world vehicle critical conditions always consist of severe crashes which are impractical to be replicated for experimenting, so we implemented the experiment on ColCOSՓ.

I co-authored a paper presenting the interesting results of this experiments and submitted it to Frontiers in Robotics and AI during the secondment in February 2020.

Besides this, I also attended the Convention on Exchange of Overseas Talent (OCS2020) and interviewed by Guangzhou TV. In the interview, I said that as a PhD that obtained the degree from abroad, what kind of career I want and what kind of support should be provided by the government.

See Xuelong being interviewed at the 1:32 mark:

I had a really great experience with my colleagues during the secondment.

Thank you for the support from the ULTRACEPT project which supported my secondment which benefited me a lot.