Project Review 22/6/2017 - project overview

Editor's Notes

• #22: Enable scalable and cost-effective documentation of archaeological findings. Enable digitization of current catalogues (by automatic processes), including the conversion from raster to digital conceptual models of the artefact classes defined in the catalogues. Enable on-the-field support to the archaeologist in the interpretation phase. Enable the archival of all data produced on a remote archive, supporting also access through advanced presentation tools. Enable reuse of all data collected and preserved in the open data repository Evaluating the impact in real conditions Establishing the conditions for a successful adoption and exploitation of project outcomes
• #37: The first step toward the achievement of this objective has been the realisation of the ArchAIDE system Mock-up, designed for working in real excavation condition. Main steps for evaluating the impact in real working conditions will start on M22 with the beginning of WP8.  
• #41: the quantity and quality of data requested by WP6 must be wider than estimated. We reach results on catalogues population but they are not enough for training the algorithms the work on the medieval and postmedieval pottery, which is of fundamental importance for developing the appearance based similarity training in WP6, was more complex than expected
• #42: the archaeological partners found new archaeological excavations from which new photomaterial can be obtained. However, the agreements with the Principal Investigators of these excavations taken over the last few months have led to an increase in time. UB start a PhD on ArchAIDE We reach results on catalogues population but they are not enough for training the algorithms Deliverable 6.1 has not been achieved on the due date. The deviation is due to the lack of training data. we propose an extension of WP5 to month 28. The date of the deliverable of WP5 will not be affected by this extension as well as it will not affect the work on WP7 and 8.
• #43: o The problem is related to the state of the art in the medieval and postmedieval archaeological pottery studies. In order to determine which pieces of ceramics could be considered as belonging to the same class, there must be some definition on when is the appearance difference is actually acceptable (and is considered as the same class) and when it’s too big (and therefore considered of different classes). However, in recent discussions it was raised that there’s no archaeological consensus on how to measure similarity as:  in some cases, it’s defined mainly by colour (and patterns may change);  in others it’s defined by the pattern;  and more. For this reason, it’s currently not possible to develop a similarity algorithm – since there’s no definition of what is considered similar; o Lacking data: as described in the work package “the similarity would be learned from examples that are tagged as similar or not-similar”. However, this requires many samples per class, and currently some catalogues only present one canonical image per class, thus preventing learning the similarity. therefore, it is not possible to train neural network classifiers on real pottery data;
• #44: o The data are unlabelled – a critical component for training a machine learning algorithm is to actually have the input data classified and labelled so that we can train the algorithm to recognize similar images as belonging to this class. Unfortunately, the current data is not labelled in any automatically consumable way; annotations are sometimes included in big paragraphs of text, but not in a machine consumable manner; o there is no unified archaeological definition for “appearance-based similarity” – In order to determine which pieces of ceramics could be considered as belonging to the same class, there must be some definition on when is the appearance difference is actually acceptable (and is considered as the same class) and when it’s too big (and therefore considered of different classes). However, in recent discussions it was raised that there’s no archaeological consensus on how to measure similarity as:  in some cases, it’s defined mainly by colour (and patterns may change);  in others it’s defined by the pattern;  and more. For this reason, it’s currently not possible to develop a similarity algorithm – since there’s no definition of what is considered similar; o Lacking data: as described in the work package “the similarity would be learned from examples that are tagged as similar or not-similar”. However, this requires many samples per class, and currently some catalogues only present one canonical image per class, thus preventing learning the similarity. therefore, it is not possible to train neural network classifiers on real pottery data;
• #45: the project logo has been created; different possibility for trademarking it have been studied; UNIPI is the only partner interested in trademarking it; decision of postponing the trademarking until the applicable commercial categories will be clearer and the Exploitation Plan is more defined
• #46: creation of the newsletter has been postponed; it has been decided to use a faster and friendly way of communication through social media it has been created the "Associates” page on the website, as a different method to communicate and involve the audience. Useremo gli associates E le mailing list dei partners
• #47: o
• #49: Being a draft, we have to say that not all partners has been able to provide the financial data updated to the end of the reporting period on month 12th, because not all the administrative offices are used to recorde the financial data within ten days after the end of the worked month. All partners worked as hard as possible to fill in the form with the more updated data as possible. As agreed with the PO the updated data will be submitted by 15th July. You can see here the periodic summary financial statement for the consortium regarding the first reporting period divided for categories of cost.
• #50: Overall, the total amount spent in this period is equal to 27,61% of the total amount and it’s consistent with the activities carried out in the period
• #51: In the first reporting period the consortium has spent the 33,24% of the planned Direct personnel costs declared as actual costs and the 15,69% of the planned direct personnel costs declared as unit costs, used by the Spanish Enterprises, Baraka and Elements. Subcontracting costs have not yet been used. This category of cost was introduced by the amendment procedure after the European Commission decision to not recognize as Personnel costs the Italian contracts called “Assegni di ricerca”, used by University of Pisa. Now, given the changes of the Annotated model Grant Agreement, these costs has been reported as “direct personnel cost”.
• #52: In the first reporting period, the higher costs were the personnel costs to an amount of 93% (89 and 4 as unit costs) and the 7% to other direct costs.
• #53: As for the other direct costs, 54% are travel costs covering the travels of the partners to attend 4 meetings (the kick-off meeting in Pisa; a technical general meeting in Barcelona; a technical meeting and a general meeting in Pisa), conferences and communication events, and travels for contacting museums and excavation teams. The 36% of other good and services has covered the promotional kit, the organisation of the meetings and the first part of the video communication. 10% has been dedicated to Tel Aviv university’s equipment.
• #54: Here you can see the categories of cost divided for each beneficiary. In this first year University of Pisa and CNR were the most involved; Baraka and Elements the less ones, they will have more efforts when the WP8 will start on month 22.
• #55: Here you can see the use of resources and the planned effort for each WP. The use of resources is in compliance with the activities carried out.
• #56: Finally, the total funding amount used for dissemination and communication activities is equal to 13,58% of the total costs spent in the first reporting period and the personnel costs are equal to 16.05% of the personnel costs involved in the 1° reporting period. The use of resources was greater for communication activities than dissemination activities, given fact that has been created the whole communication structure (internal and public) and the dissemination of the results will be more intense in the second and third year.