 I am from the University of Barcelona, and Francesca Anitina and Katie Green are presenting with me this archived project and some of the results we are achieving, but of course there are a lot of people involved in this project, I will later tell you a little bit more about it. Basically we will make an overview of the project, we will speak about the objectives that we want to arrive and we will speak a little bit about what we think are the real users needs right now. We will also speak about the resonance of what we have done with Cataloast, our reference database and what we are doing with the appearance recognition and safe recognition. We will also introduce you to the figure of our associates, that's the people that are collaborating with the archive. So every day archaeologists from around the world are working to discover and tell stories around objects from the past, investing considerable time, effort and funding to identify and characterize individual finds. It's known that around the 20 and 90 percent of the time and energy that archaeologists are spent in the classification of excavation finds and it is also known that around the 80 and 90 percent of the finds are always pottery shirts. From the Neolithic pottery was used for a lot of utilitarian purposes. In addition to this pottery is indestructible, it breaks but does not disappear like perishable goods. Consequently pottery shirts are the index posted for archaeologists being a fundamental importance for the comprehension and dating of archaeological context and for understanding the dynamics of production, trade flows and social interactions. The four pottery is for the archaeologists an extraordinary window open on the past. Today the characterization and classification of ceramics is carried out manually through the expertise of specialists and the use of analog catalogs held in our shifts and libraries. Unfortunately it is a very time-consuming activity since it's heavily dependent on human inspection and interpretation both for academic researches and professional archaeologists. Catalogs are also many times fragmented and incoherent so consultations along and fatiguing also when it's available our furnishing library. So typological classification of pottery is mainly divided into two different steps. First we analyze the shirt for the identification of the ceramic glass, the specialist look at the surface treatment, the decoration and the fabric. Then comes time for the identification of the form type. For this case we need to look at the ceramic glass paper catalogs for the specific form, analysis the section of the pot shirt and its profile and make a comparison with the published vessels. It means going through hundreds of pages and drawings. The goal of our guide is to optimize and economize this process, making no lid accessible wherever archaeologists are working. We want to revolutionize the archaeological practice, introducing a modern computer-aid approach, but we want to keep as much as possible on change the overall mythology to ensure easy adaptation and impact in the archaeological domain. This means Archi does not want to change current archaeological mythology, but the approach to study pot shirts. But it wants also to help archaeologists in their daily work using the same mythological approach, partially automatic. So Archi, maybe you have heard about us in Facebook and Instagram and Twitter because we are really active and we have also a very nice website. It's funded by European Union's everything 2020 research and innovation program and aims to create a new system for the automatic recognition of archaeological pottery from excavations around the world. It started at June 2016 and we will end on May 2019. This project involves more than 35 researchers, computer scientists, designers and video makers from universities, public research centers and private companies from different countries. Archi will support the classification and interpretation work of archaeologists and we want it to be functional also in fieldwork, post-discovery and also in laboratory situations and in museums with this innovative app designed for tablets and smartphones. Then to change the global practice of archaeology is one of our objectives thanks to the latest automatic image recognition technologies. So how does it work? You have to imagine an archaeologist at the field that finds a pottery shirt. Then the archaeologist will take the frame and will photograph it and their characteristics will be sent to a comparative collection which will activate the automatic object recognition system resulting in a response with all relevant information linked and ultimate store within a database that allows each new discovery to be shared online. In order to prove an optimal fine product for all our potential users, testbed cases are going on. The testbed cases are necessary for verifying and consolidating our vision over the required system functionalities and for performing a solid assessment of the design tool according to real archaeological requirements. Consequently, we decided to select real archaeological investigations covering both prestigious heritage sites and development excavations and surveys rather than small test cases specifically planned for the purpose of the project. The rationale is to perform the assessment in real-world conditions and not following a controlled lab experiment style in order to identify some typical interaction to scenario according to the potential use. That's why we are collaborating with different archives like here in Barcelona, different museums are also with small enterprises around Spain and Italy and with many people that are collaborating with us. I forgot to pass this one. Okay, sorry. So what we have done in this three years project, of course, was not possible to include all the pottery that exists in the archaeological record. Something that we needed to do is to choose which pottery was going to be inside the archive. Consequently, the catalogs that we needed to choose in order to train the system were directly linked to the types of pottery that we were choosing to during these three years. So finally, after a long discussion, we decided that amphora types were a good option. We are talking about amphora Roman amphora that comes from the late first century until the seventh century AD and terrasigilata. We are speaking about terrasigilata Italica, Hispanic, and South Galicia. In this case, we are not taking into account African. And in this way, we are going to be able to have in one hand, a very big frame so that you can find when the amphora are breaking down. But also we are going to test this system with fine tableware such as terrasigilata. In the other hand, we decided also to include Majolica pottery. And we are taking into account Barcelona, Majolica, Valencia, Majolica, and also Montelupo. This choice permits to create a first consistent and helpful data pool for archaeologists. The choice of these classes started from the opportunity to evaluate different types of ceramic shells. And you will see that of course, we are also choosing maybe some of you think we are going directly to choose the most easy ones because of course, for terrasigilata amphora, there is a much, there is a really clear catalogs, some of them are really well structured. But in this case, I must say that we found that terrasigilata Hispanic is not that well structured. There are a lot of people doing different things. And it has been really hard to find all the information that put it down in the database. So we are almost created a new catalog. In the case of terrasigilata italica, Conspectus is the most following, most of them only follow the catalog. So it was really easy. We were really happy to have a really structured catalog. In our first step, of course, our ICT colleagues, the first thing they did for us was to implement tools to digitize these catalogs. And in the case of Conspectus was working really good because as I said, it was really well structured. And this digitalization process allows us to populate the database. In many cases, it was very easy to do it in an automatic way. But as I said, because not all the catalogs are as well structured as Conspectus, we also needed to have some tools for manual text digitization. So the drawings that come out and the catalogs are really important for us because they allow us to create 3D models that this helps us to have a really complete database because you can search for the different types. And you will also be able to visualize these nice 3D models. As you can see, but the extraction of the drawings that you can see that different color lines, each color is given some information, like the red one will be the inside part of the shirt. The green one will be the outside. So in destruction of the drawings, we are keeping information that is relevant for the put share and that will be really necessary in the shared recognition process. So when we are speaking about what is our database, it is composed of two entities. One thing is the reference database and the other thing is the result database. The reference database contains a number of digital and digitalized catalogs, as we said, and they will be able to, and this is the core of our system. And at the end of the project cycle will form according to the static resource. The result database is extended to form a dynamic user-driving dataset for incorporation based on field and laboratory investigative and reporting workflows. The results database is extended to facilitate the capture of user-generated text and images in the field or laboratory. These are then run through the application, which is using the reference database and can generate a diagnostic of its share. This individual diagnostic fit into a larger result dataset which record the classes and types for each assemblage. So if this is the aspect of our database, as you see, you have some text for each type. You have some images. Sometimes you can also provide some few section images, the drawings, the 3D models. We have also these nice mappings where you can see where the type was manufactured. The visualization of it is really friendly and really complete. We have to say that we are also able to have multilingual vocabularies that allow a linguistic mapping. In this way, different recording traditions may not only use different words, but different levels of granularity. Quite often, not much between very specific terms, just across countries. So the mapping to a T-neutral terms allow us to within searches outside of direct stream matching. So how it works, that's what everybody's asking us. Okay, first of all, we have two different things that we have to explain. One is related to the image recognition, and the other one is related to the shape appearance. So the first step, our aim was to work with appearance-based recognition, not only for decoration, but also with the stamps. By the moment, we decided to leave the stamp-based recognition apart. We are working mainly with the decoration. Now, if you have been understanding, you have been able to see how it's working for Montalupo decoration recognition. It's some images of how the app looks like right now. As you see, you will be able to take a photo of the pot shared, and the system will give you different answers up to five, and you will be able to click to see and compare the photo catalog in order to see if you are okay with the answer, and you will need to validate it. So the system is not directly classified into the pottery, you add in some skills in order to validate with the result that you think is the correct one. So in the shape-based recognition, as explained, all the digital catalog was needed in order to extract all the profiles of all the different typologies. The system was trained with synthetic data and also with huge database with different photos that we've done all these years. The app will work in the following way. You will take the photo of the share in a correct orientation with a scale reference, and with your finger, you will draw the profile in the photo. This profile will be the one that will be analyzed by the system and will give you the different possible answers. So we assumed archaeologists will use the app to take pictures, pot shares, and they will edit the non-attributes, classify and use matching tools provided by the archive, and finally add further information if it's appropriate. Using the archive application, the user will be able to access to all his or her data because every user will have a personal space where all the classification will be stored. It will be almost helping you to write your final report. We have to say that archaeologists participate in the open research data pilot. The data created will be preserved and disseminated online and made freely for use and reused. At the end of the project, all data produced by the archive and a subset of the data from the results database will be available within the Archaeological Data Service Digital Archive. We have to say that all the information of the catalogue that we've been digitalists are not going, some of them have some copyright issues and now we are working really hard in order to have agreement with all the editors, and of course all the data coming from ADS will be open. While we are trying to have all these agreements, if we were not, if we finally were not able to be in agreement with some editors, probably you will be able to have the recognition system, but all the archaeological information behind you will need to pay it or will not be able to, the system will not be able to show it. And finally, during this year, we created the associated figure. A lot of people have been interested in the project and wanted to collaborate with us. It has been a really nice network that we built because people were sending us photos to generate all this huge data set that we needed to train the system. And hopefully we, our wish is that archive will be alive next year when the project ends and everybody will like to carry on and to create more catalogs and make archive bigger because we really think this is the future. So if you have any questions I'm afraid I will leave running to the wedding I have right now. And my colleagues Francesca and Katie will be here to answer all of your questions. In fact, I think, I don't know if I write Francesca, you brought some shares maybe, finally. So maybe you could check the app. I don't know if all of you have been already in our stand. We have a lot of people coming around asking and testing the app. And I hope I was not that long and messy.