Consultant on image sensing AI
Result of Service
The consultant will develop a detailed monitoring and evaluation plan that outlines the key milestones and indicators of their accomplishments. Under the supervision of the Chief of Section (SAS) and the project officer, the consultant will undertake the following tasks:
ESCAP will provide the geographical scope of the study areas to the consultant at the start of the assignment. It currently includes the following four local governments in South East Asia: Surabaya, Indonesia; Kuala Lumpur, Malaysia; Nakhon Si Thammarat, Thailand; and Da Nang, Viet Nam. Following a literature review/landscape analysis of the existing techniques available, the consultant will develop a methodology using image sensing algorithms to analyze photos and images of plastic litter in water bodies captured by mobile phones, cameras and other instruments on the ground and from aerial devices (such as drones). These image sensing algorithms will be based on intelligent sensing with machine learning and deep learning, including image-filtering process with “domain-boundary" analysis for color, shape, and location variables. The main feature of the methodology of intelligent sensing with AI is the analyzing function for multiple patterns of plastic waste pollution along the rivers, shore boarders and coasts based on heterogeneous data resources such as images, texts, CSV, sound, and video. When the users set up conditions and query, the algorithm will evaluate all the connected image data, by means of spatial, temporal and semantic computation functions, and integrate these results of dynamic multi-contextual computation onto chronologically-ordered maps. The consultant will also provide a series of validation and accuracy tests of image sensing algorithms according to the number of target images for the learning process.
Based on the images captured by mobile phone, aerial devices, camera and 360-degree camera in four pilot cities (>500 photos for each city) to be provided by supplier of DT and local partners, the consultant will develop geographical data visualized machine learning and deep learning models for automatic detection of marine plastic litter-photos with timestamp and GPS. This will include multi-dimensional processing and memorization with semantic computing with photo and sensing data, which will support automatic posting/mapping of macro plastic litter, and estimation of volume and flow of the plastic waste in pilot cities. In addition, using QGIS the model can overlay the geographical information data, such as polygon data of landform, location of household/factories/events and location of district/road/river etc., and statistical data of population, income-level, house index and historical data such as city/district borders. This consultant will collaborate with project partners to integrate the user-friendly model into the DT that will visualize and map plastic waste flows.
Develop online training modules for using the image sensing AI algorithms and models for participants from pilot cities, with a focus on the technicians who will continue to implement the DT in the future, to process mapping of plastic litter, and estimation of volume and monitoring the flow of the plastic waste in pilot cities. A cloud-based version of the model will be developed and available to ensure access and future usability by training participants. The consultant will also administer a survey questionnaire to the participants of the consultative and training workshops to assess the level of knowledge, awareness and use capacity of the intelligent sensing with AI model. The model will lend itself to this training process as it will be easy to use, intuitive, easy to involve local people. The training workshop will offer lectures and practical work in groups based on topical pillars and provides not only the manuals of tools to analyze the real data, but also help the participants to analyze computed results in context with visualization for expressing local environmental and social phenomena, and influences. In addition, key recommendations and explanations will be included in modules to inform decision making at the policy level. The module will need to be developed by December 2020 in time for ESCAP Committee on Environment and Development.
Provide necessary participation and support on international fora, workshops, seminars and trainings, including Committee on Environment Development 2020, the 3rd session of Committee on ICT&STI (VC), two virtual meetings/trainings related to the project, participation of the Asia-Pacific Forum on Sustainable Development in Bangkok, Thailand.
1 July 2020 to 31 January 2021 (7 months)
Duties and Responsibilities
Space Application Section (SAS) of the Information and Communications Technology and Disaster Risk Reduction Division (IDD) is supporting the implementation of the project “Closing the Loop: Scaling up Innovation to Tackle Marine Plastic Pollution in ASEAN Cities", which aims to reduce the environmental impact on cities in South East Asia of plastic waste pollution and leakage into the marine environment. The project will develop a customised, open source, innovative digital tool (DT) tailored to the needs and capacities of four local governments in South East Asia to monitor and visualize plastic waste leakage with a view to: i) prevent this waste from entering water streams and the ocean; and ii) improve identification of waste generation hot spots and improve their management at the municipal level. The project is expected to be completed in March 2021.
In this regard, SAS plans to recruit a consultant with good knowledge and experience on Intelligent Sensing with AI (machine learning and deep learning) for macro plastic litter in water bodies, spatial statistics and related GIS-based programing, applications and training to develop a methodology and AI algorithms to enhance image sensing. This methodology will be complementary to the DT mentioned above. In addition, a series of e-learning modules will be developed based on the AI methodology and the supplier-developed DT.
Competency: Good competency and knowledge on innovative solutions for emerging regional needs and trends in geospatial information applications in South East Asia, and good knowledge and experience on intelligent sensing with AI (machine learning and deep learning) for plastic-garbage, spatial statistics and related GIS-based programing, applications and training to conduct the above-mentioned assignments, as supplementary of the digital tools on enhancing image sensing AI and training. Capacity in developing knowledge products and operational tools on geospatial information applications will be required and teaching or delivering trainings.
Academic Qualifications: Advanced degree (or equivalent) related to technology-related areas (preferably database, data mining, image processing, semantic computing, AI, machine learning, remote sensing and GIS).
Experience: At least 10 years experiences in image processing, geospatial information applications and/or teaching/training in relevant areas. Work experience on preparation of training materials and organization of international and regional events and trainings.
Language: Good capacity of English speaking and writing skills.