ICLR 2020 Workshop on
Computer Vision for Agriculture (CV4A)

ICLR 2020 Workshop on Computer Vision for Agriculture
April 26th 2020

Artificial intelligence has invaded the agriculture field during the last few years. From automatic crop monitoring via drones, smart agricultural equipment, food security and camera-powered apps assisting farmers to satellite imagery based global crop disease prediction and tracking, computer vision has been a ubiquitous tool. This workshop aims to expose the fascinating progress and unsolved problems of computational agriculture to the AI research community. It is jointly organized by AI and computational agriculture researchers and has the support of CGIAR, a global partnership that unites international organizations engaged in agricultural research for a food-secure future.

Virtual Workshop

Due to the COVID-19 pandemic, ICLR 2020 will be hosted as a fully virtual conference. Our workshop will also be virtual, and we have a full day of events on Sunday the 26th of April, as well as networking events and socials throughout the week on our "Virtual Booth".

Note that you will need to be registered for the ICLR conference to participate in the online workshop.

Everyone, however can watch it live here, on our website.

• You can find the detailed program here, and out virtual booth events here.
• Registered participants can access our workshop livestream and chat discussion here.
• For each session we have a live discussion with the speaker(s). To join the discussion and ask questions, you need to join our Zoom webinar session. To do that click on “Live” inside the top banner of our workshop livestream page.

Watch Live

If the embedded video doesn't show above, please try clearing the cache for this website.

You can also watch the CV4A workshop livestream from the slidelive website, at this page.

Virtual Booth Events

We are excited to announce that the following events will take place on our workshop's "Virtual Booth" during the week of the conference:

• April 27th, 8:00 AM (EST): Shaping a New Fund for Labeled Data

  The Rockefeller Foundation, in concert with partners in across industry and philanthropy, will launch a new fund for the creation, expansion, and maintenance of equitably labeled datasets that enable the robust application of machine learning tools. The Foundation expects the first call for proposals will focus on agricultural imagery and would like to get feedback on the concept and proposed implementation from the CV4A community.

  • Register for free to attend this event.

• April 27th, 10:00 AM (EST): Agriculture for Computer Vision: Bridging Application and Research

  A conversation around how challenges in agriculture shape our broader understanding of computer vision and vice versa. Computer Vision has tremendous capacity to improve agriculture and address key societal needs. Additionally, agriculture poses new challenges and opportunities for machine learning and computer vision. This networking session brings together researchers in both academia and industry to discuss some of the key technical challenges and hopefully foster future collaborations. Event organized by Intellinair.

• Register for free to attend this event.
• April 28th, 13:00 - 19:00 (UTC/GMT): Agriculture, Forestry, and Other Land Use (AFOLU) Day (In collaboration with the Climate Change AI workshop)

  On Tuesday we join forces with the with the Climate Change AI Workshop for a series of events in the intersection of AI, agriculture and climate change.

  More details can be found here.

  • Register for free to attend this event.

  Click here for the Zoom meeting

Workshop Program

Below you can find the program in the following forms:

• Online version of the virtual workshop program
• Online version of the Virtual Booth event schedule
• Google calendar version, including the Virtual Booth events

Program at a glance

Tip: Click on any session to see more details!

Workshop Session Details

---

Opening remarks from the organizers

Speaker: Yannis Kalantidis on behalf of the organizers.
Slides:

Back to the program at a glance.

---

---

Invited Talk:
One Acre Fund: Helping Smallholder Farmers

Speaker: Emiel Veersma (One Acre Fund)
Short bio: Working as Data Scientist at One Acre Fund, I'm responsible for supporting the organization with analyzing the vast amounts of data. This involves supporting farmers with planting and fertilizer advice, analyzing client payment data and setting up experiments to optimize the pricing strategy.
Slides: Link to slides.
Talk Abstract: One Acre Fund is a nonprofit organization that supplies smallholder farmers in East Africa with asset-based financing and agriculture training services to reduce hunger and poverty. Using a market-based approach, One Acre Fund facilitates activities and transactions at various levels of the farming value chain, including seed sourcing and market support. During the talk, I'll talk about Data Science projects we've been working on at One Acre Fund, and what our next steps will be.

Video:

Back to the program at a glance.

---

---

Invited Talk:
Digital Farm Twinning in Africa

Speaker: Mohamed Akram Zaytar (Faculté des Sciences et Techniques de Tanger, Morocco)
Short bio: Mohamed Akram Zaytar is a Ph.D. candidate at the Faculty of Sciences and Technologies Tangier, doing work at the intersection of machine learning and environmental science. His research focuses on weather forecasting, air quality monitoring, and precision agriculture with remote sensing data.
Slides: Link to slides.
Talk Abstract: African smallholder farmers face considerable difficulties in coping up with the unpredictable nature of farming conditions (e.g., environment, weather & soil). For them, making uninformed decisions often decreases the productivity of their farms and hence their livelihoods. In this presentation, we go over our past, current, and future research efforts that aim at aiding smallholder farmers in Africa. Using a range of data types contextualized to form digital twins of farms, we present a method for recommending cultivation time-windows for plowing and a scalable approach for weakly-supervised small-scale boundary estimation.

Video:

Back to the program at a glance.

---

---

Invited Talk:
Mapping Geographical Diversity of Crop Disease Surveillance in a Crowdsourced Environment

Speaker: Joyce Nakatumba (Makerere University, Uganda)
Short bio: Joyce Nakatumba-Nabende is a Lecturer in the Department of Computer Science at the School of Computing and Informatics Technology, Makerere University. She holds a PhD in Computer Science from Eindhoven University of Technology, The Netherlands. Joyce is the head of the Makerere Artificial Intelligence research lab where she carries out research in the application of machine learning techniques in areas of agriculture and health. 

Talk Abstract: This talk will discuss work around a scaled community sensing model when open sourced technology has been deployed to acquire crop image data in a crowdsourced environment. We perform a spatial assessment on the diversity of image data collected to understand the state of crop health for different geographical regions. The talk will delve into challenges in this crowdsourced model and potential data sources, enhancements that can argument crowdsourced image data to provide a holistic pest and crop disease surveillance model in Uganda.

Slides: Link to slides.

Video:

Back to the program at a glance.

---

---

Invited Talk:
Advancing the delivery of smart digital tools to smallholder farmers in Africa - Ready or Not?

Speaker: Julius Adewopo (International Institute of Tropical Agriculture, Rwanda)
Short bio: Julius Adewopo is geodata scientist within multidisciplinary background in soil science and digital agronomy. In the past 5 years, he has led several projects and initiatives, mainly focused on delivery of digital tools and analytics for decision-support in smallholder agricultural systems, including smart ICT surveillance system for banana disease in East Africa, [near-]real-time crowdsourcing of food prices in Nigeria, and bridging of ground-based weather data gap in Africa. Last year, he was honored by the Agropolis Foundation as the recipient of the Louis Malassis laureate award for young promising scientist, and he continues to advance collaboration on various themes to address most-pressing challenges for agricultural production in Africa. 

Talk Abstract: Digital technologies are increasingly accessible in Africa and the agricultural development sector embraces their potential. Projects adopting digital tools in their interventions often promise unprecedented outcomes and impact. Yet, the understanding farmers’ readiness to adopt digital innovations is limited and national statistics of access to smart technologies often tell a skewed story. For instance, recently, a robust survey of 700 farmers in Rwanda shows that only 3% possess or use smartphones, yet, many interventions assume that these farmers will uptake tools that are deployed on smartphone-based platforms. Further, farmers (as tool users) are often neglected in smart tool development and their capacity to use the tool is seldom considered. This presentation will highlight some of the nuances of co-developing smart digital tools based on multi-project experiences in sub-Saharan Africa, and raise important questions (and insights) for envisioned sustainability of digital agriculture within smallholder farming systems.

Slides: Link to slides.

Video:

Back to the program at a glance.

---

---

Paper session I:
Spotlights (Europe/Africa/Asia-based presenters)

The following papers are going to be presented during this session, in the order listed below.

• A Deep Learning Approach for Determining Effects of Tuta Absoluta in Tomato Plants [pdf]

  Authors: Denis P.Rubanga, Loyani K. Loyani, Mgaya Richard, Sawahiko Shimada

  Video:
  

  
  

• Fine grained classification for multi-source land cover mapping [pdf]

  Authors: Yawogan Jean Eudes Gbodjo, Dino Ienco, Louise Leroux, Roberto Interdonato, Raffaelle Gaetano

  Video:
  

  
  

• Evaluation of Model Selection for Kernel Fragment Recognition in Corn Silage [pdf]

  Authors: Christoffer Bøgelund Rasmussen, Thomas B. Moeslund

  Video:
  

  
  

• Estimating Grape Yield on the Vine from Multiple Images [pdf]

  Authors: Daniel L. Silver, Jabun Nasa

  Video:
  

  
  

• Early Disease Diagnosis for Rice Crop [pdf, poster]

  Authors: M. Hammad Masood, Habiba Saim, Murtaza Taj, Mian M. Awais

  Video:
  

  
  

• Hyper-spectral NIR and MIR data and optimal wavebands for detection of apple tree diseases [pdf]

  Authors: Dmitrii Shadrin Mariia Pukalchik Anastasia Uryasheva Nikita Rodichenko Dzmitry Tsetserukou

  Video:
  

  
  

• Image-based phenotyping of diverse Rice (Oryza Sativa L.) Genotypes [pdf]

  Authors: Mukesh Kumar Vishal, Dipesh Tamboli, Abhijeet Patil, Rohit Saluja, Biplab Banerjee, Amit Sethi, Dhandapani Raju, Sudhir Kumar, R N Sahoo, Viswanathan Chinnusamy, J Adinarayana

  Video:
  

  
  

• A single image deep learning approach to restoration of corrupted remote sensing products [pdf]

  Authors: Anna Petrovskaia, Raghavendra B. Jana, Ivan V. Oseledets

  Video:
  

  
  

Back to the program at a glance.

---

---

Paper session II:
Oral Presentations (Europe/Africa/Asia-based presenters)

The following papers are going to be presented during this session, in the order listed below:

• Transfer Learning of Photometric Phenotypes in Agriculture Using Metadata [pdf]

  Authors: Dan Halbersberg, Aharon Bar Hillel, Shon Mendelson, Daniel Koster, Lena Karol, Boaz Lerner

  Video:
  

  
  

• Mobile-Based Deep Learning Models for Banana Diseases Detection [pdf]

  Authors: Sophia Sanga, Victor Mero, Dina Machuve, Davis Mwanganda

  Video:
  

  
  

Bayesian aggregation improves traditional single image crop classification approaches [pdf]

Authors: Ivan Matvienko, Mikhail Gasanov, Anna Petrovskaia, Raghavendra Belur Jana, Maria Pukalchik, Ivan Oseledets

Video:

Back to the program at a glance.

---

---

Invited Talk:
Integrating Earth Observations in National Agriculture

Speaker: Catherine Lilian Nakalembe (University of Maryland, College Park)
Short bio: Dr. Nakalembe is an Assistant Research Professor at the University of Maryland. She is the NASA Harvest Africa Program Lead, a member of the NASA SERVIR Applied Sciences Team and serves as the Agriculture and Food Security Thematic Lead. She has broad interest ranging from agriculture remote sensing, food security to climate change and supports several capacity bundling actives in the context of the GEO Global Agricultural Monitoring Initiative. 

Talk Abstract: For many African countries, food security is one of, if not the most, pressing issues of today. My job as Africa Program Lead for NASA Harvest is to help countries build their own agricultural monitoring systems based on free and low-cost satellite data to inform life-saving decisions related to food security sooner and with a deeper evidence-base. Our work at NASA Harvest focuses on using satellite data to understand agriculture and food security at local to global scales. Satellite data have been used in agriculture for decades by organizations like USDA, but not in many other countries especially in Africa. NASA’s Applied sciences programs Harvest and SERVIR are redoubling NASA’s investment in making satellite data more useful for agriculture monitoring. Combined with field data, satellite data help us understand sooner, and at a larger scale, what food crops are growing where, how they are doing, estimate how much food will be produced in a season, and provide early warnings of crop failure. This early warning gives organizations more lead time to prepare responses and even mitigate impacts altogether.

Video:

Back to the program at a glance.

---

---

Invited Talk:
Wheat Rust Early Warning Systems in Ethiopia – Using New Technologies to Combat Crop Disease

Speaker: David Hodson (CIMMYT International Maize and Wheat Improvement Center)
Short bio: David Hodson is a Principal Scientist with CIMMYT bases in Mexico. He has over 20 years of experience executing and managing GIS-related projects and programs for agricultural research and development in developing countries. For the last 10 years he has worked on developing and coordinating a Global Wheat Rust Monitoring System in response to the threat posed by wheat stem rust Ug99. The wheat rust monitoring system now covers approximately 40 countries in Africa, Asia and Latin America. In addition, Hodson is involved in projects using advanced modeling for wheat rust early warning, and also on improved pathogen and host diagnostics using molecular tools. His research focuses on the surveillance and monitoring of emerging cereal disease threats and the application of geo-spatial technology for improved decision support. 

Talk Abstract: Wheat rusts pose a major threat to food security in Ethiopia, with several devastating epidemics in recent history. To help prevent major disease outbreaks, early detection and timely control are essential. In response to the wheat rust problem in Ethiopia a consortium of national and international partners have created one of the most advanced, operational crop disease early warning and advisory systems in the world. This early warning system includes several advanced technologies and operates in near real-time within the wheat season. Key elements include; near real-time field and mobile phone surveillance data, mobile nanopore sequencing diagnostics, meteorologically-driven spore dispersal, disease environmental suitability forecasting and a platform for timely communication to policy-makers, extension agents and small-holder farmers. The existing early warning system and the planned development of additional, new components for enhancement based on machine learning and remote sensing will be described.

Video:

Back to the program at a glance.

---

---

Invited Talk:
Technology Through the eyes of a farmer

Speaker: Noah Nasiali Kadima (AFarmers & Africa Farmers, Kenya)
Short bio: Multi-Award Winning Farmer, Community Entrepreneur, & Agripreneur 

Talk Abstract: Technology has been developed to solve the worlds problems. Most of the technology developed works well in the sectors that it has been designed. In the farming/agriculture sector this has also been done with great results but why are farmers not using the technology that has been developed? At what point should we engage with the farmers and develop solutions with them and not for them?

Back to the program at a glance.

---

---

Live Panel Session: Panel on Food Security

Moderator: Jingying Yang (Partnership on AI)
Speakers:

• Denis Fidalis Mujibi (USOMI Agriculture)
• Sarah Papazoglakis
• Amanda Ramcharan
• Hannah Kerner (University of Maryland)
• Christopher Abiodun (Farmcrowdy)

---








---

CV4A Challenges session

• CV4A Challenges overview presentation Celina Lee (Zindi.africa)

  Video:
  

  
  

• The CGIAR Wheat Rust Detection Challenge

  Crop disease has threatened the global food supply since the beginning days of agriculture, over 10,000 years ago. Wheat, along with maize and rice, represent the three most important crops to humanity. One of the most devastating plant disease is a fungal infection of wheat, called rust. A particularly virulent strain of wheat rust known as Ug99 has emerged from Uganda, with up to 100% yield losses from infected fields. Ug99 has spread across Africa and is moving into South Asia, threatening regional food security. The international agricultural community is focusing on methods to monitor and track the presence and spread of Ug99 across the world. Different strains of wheat rust can be initially identified by visual symptoms with the naked eye. The challenge is to develop a computer vision-based (CV) detection solution for differentiating strains of wheat rust which can be deployed around the world. If accurate CV models can be generated, these models can be deployed across the network of wheat farmers, breeders, and pathologists around the world to track the emergence and spread of wheat rust, identify promising resistant varieties, and to disseminate resistant varieties to millions of farmers all over the world.

  This challenge is organized and supported by CGIAR and CIMMYT.

  • Challenge overview - David Guerena (CIAT)

    Video:
    

    
    

  Presentations from the award winners:

  • 1st place overall: Lukas Picek & Miroslav Valan [slides]

    Video:
    

    
    

  • 1st place African citizen: Karim Amer

    Video:
    

    
    

  • 1st place female-identified African citizen: Aisha Alaagib and Jeremiah Fadugba

    Video:
    

    
    

• The Radiant Earth Foundation Crop Classification using Earth Observations Challenge

  Earth Observations provide invaluable data across different spatio-temporal scales enabling applications for agricultural monitoring. Meanwhile, machine learning (ML) techniques can be utilized to advance these applications, and develop faster, more efficient and scalable models. In this challenge, a training dataset of crop types in Kenya will be provided and participants are asked to develop the best model to predict crop types using mutli-spectral and radar data from Sentinel-2 and Sentinel-1 satellites. The data includes a crop type label as well as time series of measurements from each of the satellites during the growing season. Training dataset for this challenge is developed from ground reference data collected by Plant Village team in the year 2019. Radiant Earth Foundation in partnership with Plant Village has generated the training data, that will be hosted on Radiant MLHub with an API that all participants can easily access.

  This challenge is organized and supported by the Radiant Earth Foundation and Plant Village.

  • Challenge overview - Hamed Alemohammad (Radiant Earth Foundation)

    Video:
    

    
    

  Presentations from the award winners:

  • 1st place overall: Karim Amer

    Video:
    

    
    

  • 1st place African citizen: Tolu Sotonwa

    Video:
    

    
    

  • NAVER Labs Europe prize for 1st place female-identified African citizen: Ansem Chaieb
  -->

Back to the program at a glance.

---








---

Invited Talk:
Towards 2050: Big data driven decision support in integrated smallholder farming systems

Speaker: Denis Fidalis Mujibi (USOMI)
Short bio: Dr Denis Fidalis Mujibi (PhD) has extensive experience working with smallholder farmers in Africa. Denis’ passions include development of channels for enhanced uptake of appropriate technology in developing world farming systems. As CEO of USOMI Limited, he is leading the emergence of structured commodity markets to maximize profitability for farmers.

Talk Abstract: The main challenge facing smallholder farmers is the segmented value chains they operate in. Farmers cannot easily access inputs and services, which leads to low yields; partly due to lack of cash or credit services to buy the right inputs up front, but also fragmented grower base which is uneconomic to service. The landscape is thus riddled with unpredictable supply, most of which doesn't meet market quality and quantity and woefully low yields. In order to improve farm productivity, it is important to provide farmers with critical support including access to inputs, agronomic support, data-driven actionable feedback and mechanization. USOMI is developing a purpose-built structured trading platform that drives maximal productivity, guaranteed market access and inclusive value chain participation for smallholder farmers in Africa. This will be achieved by organizing smallholder farmers into producer groups that facilitate guaranteed access to inputs, agronomy support and mechanization services to maximize yields and production efficiency at farmer level, as well as acquisition of forward contracts by large buyers such as millers and processors, at the market end of the value chain. The innovative application of modern technologies such as weather data, hyperspectral imaging and remote sensing using satellite imagery, and data collection using soil sensors in a precision agriculture framework will ensure that yield maximization strategies are data driven and repeatable.

Back to the program at a glance.

---








---

Invited Talk:
The PlantVillage Project

Speaker: Annalyse Kehs (Penn State University)
Short bio: Annalyse Kehs is the Executive Director of PlantVillage, a global platform that uses AI, satellites and cloud computing to help smallholder farmers adapt to climate change.She worked on projects ranging from machine learning modeling on satellite imagery of fields to building a 12-person team in Kenya. In addition to working directly with farming communities in Kenya and India she has engaged leadership at the United Nations Food and Agriculture Organization in Rome on the powerful role technology can play in transforming the lives of low-income farmers struggling to cope with climate change.

Talk Abstract: PlantVillage is a non-profit, global public-good platform that delivers expert-level knowledge to smallholder farmers. We strive to form and leverage relationships with our farmers, local governments and various organizations, like UN FAO. We are currently the UN’s frontline pest/disease management tool across 66 different countries and in 29 languages. Our most important relationship is with farmers, specifically the subsistence farmers we work with and it is critical that we ensure the delivery of up-to-date, climate adaptive information to enable farmers to make the best decisions for their farms and families. We do this by looking at constraints subsistence farmers in rural areas face daily (whether pests/diseases or climate change) and provide in-field, in-season solutions using a hybrid model for free.
Slides: Link to slides.

Back to the program at a glance.

---








---

Paper session III:
Oral & Spotlight Presentations (America-based presenters)

The following papers are going to be presented during this session in the order listed below:

• Satellite-based Prediction of Forage Conditions for Livestock in Northern Kenya [pdf, blogpost]

  Authors: Andrew Hobbs, Stacey Svetlichnaya

  Video:
  

  
  

• Field-Level Crop Type Classification with k Nearest Neighbors: A Baseline for a New Kenya Smallholder Dataset [pdf]

  Authors: Hannah Kerner, Catherine Nakalembe, Inbal Becker-Reshef

  Video:
  

  
  

• Apple Defect Detection Using Deep Learning Based Object Detection For Better Post Harvest Handling [pdf]

  Authors: Paolo Valdez

  Slides:
  

Back to the program at a glance.

---




Google Calendar version

This is the easiest and most dynamic way to track our workshop sessions during the workshop day as well as all our Virtual Booth events during the week.








Workshop Proceedings

All papers presented in this workshop can be openly accessed through the arXiv platform.

People

Sponsors

Platinum

Gold

Silver

Bronze

_{Image Credit: NASA/GSFC/Jeff Schmaltz/MODIS Land Rapid Response Team}