CROMATEK pilot was prompted in response to a notorious concern in the market of high brilliance parts, and specially in a demanding industry as the automotive sector, with extremely high surface finishing requirements. Tecnalia, together with Maser, has developed an automated inspection system for chromed parts, where defects of up to 0.1mm2 or 0.1mm in the case of slim scratches, can be detected. This quality control system makes use of different technologies such as robotics, computer vision and neural networks for deep learning algorithms.
BIONIC Aircraft project has recently finished. This is the acronym of Increasing resource efficiency of aviation through implementation of ALM technology and bionic design in all stages of an aircraft life cycle, and its budget was €7.9million, being funded by the European Commission thanks to Horizon2020 program. The general goal of the project was increasing the efficiency of the resources when manufacturing airplanes, fostering the intensive use of Additive Layer Manufacturing (ALM) techniques. This technology will allow reducing emissions during the flights due to the weight saving when using complex geometry, light and high resistance parts. This kind of parts, also called Bionic parts, avoid wasting raw material because no classical machining techniques are needed for their development. In this Project, Tecnalia´s Computer Vision Team tackled a task with the aim of asset the quality control of those new bionic parts during their use, this is [...]
Deep learning diagnostic algorithms are proving comparable results with human experts in a wide variety of tasks, they still require a huge amount of well annotated data for training which is often non-affordable. Metric learning techniques have allowed a reduction on the required annotated data allowing few-shot learning. Existing deep metric learning loss functions have made possible generating models capable of tackling complex scenarios with the presence of many classes and scarcity on the number of images per class, not only work for classification tasks, but to many other clinical applications where measuring similarity is the key. Currently used state-of-the-art loss functions still suffer from slow convergence due to the selection of effective training samples that has been partially solved by the multi-class N-pair loss by simultaneously adding additional samples from the different classes. The constellation loss goes one step further by simultaneously learning distances [...]
School of AI invited Tecnalia researchers Alfonso Medela and Miguel G. San Emeterio to give a talk on “Deep Learning”. School of AI is a non-profit organization whose goal is to provide on AI all over the world. At the moment they have settled in 400+ cities distributed around 80+ countries which are managed by a group of 800+ people around the world. The newly created Bilbao branch of School of AI invited our organization's researchers to its third event held in Bilbao to share some of their knowledge on Deep Learning. The took place at the facilities of the University of Mondragon at Bilbao. Alfonso and Miguel talked about “What is Deep Learning?”, Artificial Intelligence, Neural Networks, Machine Learning and the impact, current issues and future research lines.
Continuing the work done in 2018 in Nuevo Laredo, Tecnalia researchers have participated in training tasks in several courses offered at the Technological Institute of Saltillo, Mexico. Saltillo is the capital city of the State of Coahuila, located in northern Mexico where the main economic activities are generated by the manufacturing industry, services sector and transport; contributing 3% of national GDP. The government of the State of Coahuila hired Tecnalia to be part of the HUB Coahuila Innovation 4.0, a joint initiative by the Ministry of Economy of the State, the National Chamber of the Transformation Industry Delegation Coahuila Southeast (CANACINTRA) and the Technological Institute Saltillo (ITS). It was created to support the Industry 4.0 revolution in the State of Coahuila, and to be a dynamic agent of the industry through the transfer of knowledge between companies and academia. Within the first phase of this initiative, [...]
Last April 8th-11th took place the IEEE International Symposium on Biomedical Imaging (ISBI), a scientific conference dedicated to mathematical, algorithmic, and computational aspects of biological and biomedical imaging, across all scales of observation. The member of piccolo team and Tecnalia Research & Innovation, Alfonso Medela, presented the paper “Few-shot learning in histopathological images: reducing the need of labelled data on biological datasets”. The team has been working on a few-shot approach in parallel with the acquisition of the datasets. To overcome the problem of scarce data in new imaging modalities such as OCT and MPT, few-shot techniques provide a solution to create algorithms out of a small number of images. The results showed that by using the proposed method it is possible to beat classical transfer-learning approach when only few images per class are available. The results encouraged the team to continue working on the same track and as [...]
The spatial scales (from meters to kilometers) and temporal (hours, days, years) that characterize the coastal dynamics, make that the classical measurement techniques are limited and very expensive, to study the behavior of coastal systems. The inclusion of measurement techniques by means of video images (commonly called coastal videometry), nowadays allows to describe physical processes on a wide range of spatial and temporal scales, something unthinkable until very recently. A coastal videometry system consists in cameras installed on the coast that allow the capture of images and their spatial referencing. The products derived from the image processing give very interesting information for the different activities that are developed in the coastal areas and that depend on the waves, currents and tide (hydrodynamic conditions) as well as the configuration of the beach, dunes, channels and bars (sedimentary elements). But trying to go a step further, based on these tools, [...]
TECNALIA and the company specializing in the construction of industrial equipment Automatismos Maser have developed an automated solution for the final verification of the parts for the foundry company Betsaide. The development has been carried out with the support of the Basque Industry 4.0 program, aimed at financing intelligence projects in media and production systems and managed by the SPRI Group. The solution consists of a cell composed of two synchronized robotic arms and two camera systems. The first one captures 2D images to identify internal defects and inspect dimensional and mold variation aspects. Meanwhile, the second camera is a structured light 3D system with high processing speed for the control of surface defects, including those of curved surfaces, and aspects related to thickness in flat areas. This new system has allowed Betsaide to win a contract that, according to a note issued by Tecnalia, has given them [...]
Siamese networks were first introduced by Bromley and LeCun  in early 1990s to solve signature verification as an image matching problem. A similar Siamese architecture was independently proposed for fingerprint identification by Baldi and Chauvin  in 1992. Later in 2015, Gregory Koch et al.  proposed to use Siamese neural networks for one-shot image recognition. Siamese neural networks are designed as two twin networks that are connected by their final layer by means of a distance layer that is trained to predict whether two images belong to the same category or not. The networks that compose the siamese architecture are called twins because all the weights and biases are tied, which means that both networks are symmetric. Symmetry is important as the network should be invariant to switching the input images. Moreover, this characteristic makes the networks much faster to train since the number of [...]
Industrial pick & place applications automate processes where loading and/or unloading unitary product is needed. This kind of installations usually use robots or articulated arms to manipulate product from one area to another in stacking, box fitting and palletizing work. Both physical product characteristics (shape, deformity degree and adhesiveness) and disposition, as well as transfer velocity and manipulation zone distinctive features establish task complexity. To optimize installation efficiency and reduce cycle time, it is common using line tracking technique to take and unload product from or to a conveyor belt while it is running. This is possible with robots being commanded by an encoder signal. Another important point is related to grip design. In some cases, it is enough using simple vacuum grips provided with suction pads which allow to grab piece at desirable speed. A more complex option consists on working with grabbing clamps, pneumatic cylinders [...]