In order to boost the participation of the AI community, an Online Challenge will be organized.
The goal will be to invite the AI community interested in AI for horticulture and to motivate them in participating in the Hackathon and Greenhouse Growing Challenge.
It will not be necessary to participate in the Online Challenge to be eligible for participation in the Hackthon. However, it is advisable, since we will use the Online Challenge for scouting of talents. Participating teams will not only have a chance for a price and a wild card for the Greenhouse Growing Experiment, but also the possibility to exercise for the Hackathon.
We believe the following expertise from AI community is needed towards fully autonomous crop control: Machine learning skills and Computer vision skills. Machine learning skills will be tested in the interaction with a lettuce growing simulator. The simulator will consist of a simple greenhouse climate and crop production model that will be provided. Computer Vision skills will be tested on real lettuce images. A series of annotated images will be provided as training dataset.
These are two parts of the online challenge:
Part A : Computer vision challenge
Teams will get access to a series of lettuce plants. Images are taken with a RealSense camera under defined conditions and contain images of individual lettuce plants of different varieties in different growth stages and grown in different growing conditions. Each image is connected with information on the ground truth plant traits, such as plant diameter, plant height, plant fresh weight, plant dry weight, and leaf area. Teams use ca. 300 images provided in batches to develop a computer vision algorithm during the preparation phase. This algorithm will have to be able to estimate the plant traits of a series of ca. 50 unseen lettuce plant images provided during the Online Challenge under limited time and memory constraints. The computer vision algorithms have to detect the plant parameters described above.
Part B : Machine learning challenge
Teams will get access to a virtual simple greenhouse climate and lettuce production model (simple simulator). The simple simulator consists of a given set of outside climate conditions, a given greenhouse type and given greenhouse actuators (ventilation, heating, lighting, screening). It needs to be provided with a series of climate setpoints (ventilation strategy, heating strategy, lighting strategy, screening strategy per timestep) as inputs. The input climate setpoints will activate the available virtual actuators, which will control the inside greenhouse climate. The realised inside climate parameters will be provided as a feedback value. Since the crop growth in the simulator is determined by the realised greenhouse climate, also the crop growth parameters (fresh weight, height, diameter) over time will be provided as output. Teams will have to develop machine learning algorithms to feed the simple simulator with the optimised control parameters in order to maximise net profit. During the preparation phase teams can interact with the simple simulator for algorithm development. During the Online Challenge this algorithm should be suitable to control the growth of a virtual crop in a virtual greenhouse under changed conditions (e.g. other weather conditions, different greenhouse type, different lettuce type) and limited time constraints.
  • The Online Challenge is open to students and researchers from universities and research centres and experts, from companiesand start-ups.
  • Teams of at least 2 real persons are eligible to subscribe.
  • Team membersmust have the following expertise: 1. Machine learning 2. Computer vision. This expertise has to be demonstrated by professional or academic engagement.
  • Each team will appoint a team-captain who acts as contact person.
  • Each participantis only allowed to participate in one team and to subscribe to the Online Challenge once.
  • Participants must subscribe via the official site with their professional or educational email account, thus company/start-upor university/research centre account.
  • Names of participants and email addresses will not be disclosed by the organisers, only team names and number of teams will be public. If teams reach the top 5 ranking at the end of the Online Challenge, they automatically agree that the names of participants will be disclosed to honour them.
  • We encourage teams from different countries and continents to participate. We encourage cooperation of different experts from different start-ups/companieswith students and researchers form universitites/research centres. We encourage to engage with experts in the field of horticulture but this is not mandatory.
  • Good English language skills are required.
  • The maximum number of teams acceptedin the Online Challenge will be 200. In case of a higher number of subscriptions the organizers and jury will rank the subscriptions and select based on the criteria stated above. The jury’s decision will be final and will not be subject to debate.
  • 15 April – 20 May 2021 0:00 GMT: Register open for Teams
  • 20 May - 1 June 2021: Confirmation of admission of teams, links to images and simulator will be provided incl. more detailed information.
  • 1 June 2021: Teams will get technical information for the Online Challenge and access to the lettuce images including ground truth plant traits and access to the simple greenhouse-crop simulation model (simple simulator)
  • 1 June 0:00 – 11 July 2021 23:00 GMT: Preparation time for teams to develop their computer vision and machine learning algorithms
  • 12 July 2021 0:00 GMT: Possibility of re-training machine learning algorithm during 24 h
  • 13 July 2021 12:00-12:05 GMT: Online Challenge computer vision–teams have to provide results on unseen images
  • 13 July 2021 12:00-13:00 GMT: Online Challenge machine learning –teams have to provide results on unseen lettuce simulator
  • 14 July 2021: Online Challenge event incl. winning ceremony
Part A computer vision challenge
  • For the computer vision challenge teams will get access to a series of ca. 300 images of lettuce plants during the preparation phase.
  • Images are taken with a RealSense camera under defined conditions and contain images of individual lettuce plants of different varieties in different growth stage and grown in different growing conditions. Each image is connected with information on the ground truth plant traits, such as plant diameter, plant height, plant fresh weight, plant dry weight, and leaf area.
  • Teams use the images to develop a computer vision algorithm. This algorithm will have to be able to estimate the plant traits of a series of ca. 50 unseen lettuce plant images provided after the preparation phase under limited time and memory constraints. The computer vision algorithms have to detect the plant parameters described above.
  • For each picture and plant trait the absolute difference between estimated value (teams’ computer vision algorithm) and real value (WUR ground truth measurement) will be calculated and normalised. The total sum of absolute normalised difference will be calculated per team. The team with the lowest difference will be ranked first for this part of the Online Challenge.
    There will two ranking boards available:
    • The first is the public ranking board, which will be visable during the preparation phase of the Online Challenge. On 1 June 2021 00:00 GMT, teams will be provided with batches of images (out of the ca. 300 total images). After viewing each batch of images, the team can upload at most one new submission of the image prediction results. After every new submission, the public ranking board will show the scores of the latest submission, as well as the ranking list of all participating teams. On July 11th 2021 23:00 GMT, the public board will be closed and no new submission will be accepted anymore.
    • On July 13th 12:00 GMT, a private ranking board will be opened and a new set of 50 unseen lettuce plant images will be provided to the teams. Teams must submit the image prediction on this image set within 5 minutes. After 12:05 GMT the submitting will be rejected. Teams can only see their own scores in the private board before the announcement of the final results on July 14th.
  • A total ranking of teams will be made based on the private board score. The team with the highest score will be ranked first on the ranking board.
Part B machine learning challenge
  • For the machine learning challenge, teams will get access to a simple greenhouse climate and lettuce production model (simple simulator) during the preparation phase.
  • The simple simulator consists of a given set of outside climate conditions, a given greenhouse type and given greenhouse actuators (ventilation, heating, lighting, screening). It needs to be provided with a series of climate setpoints (ventilation strategy, heating strategy, lighting strategy, screening strategy per timestep) as inputs. The input climate setpoints will activate the available actuators, which will control the inside greenhouse climate. The realised inside climate parameters will be provided as a feed back value. Crop management consists of defining plant density (number of plants m-2) over time. Since the crop growth in the simulator is determined by the realised greenhouse climate, also the crop growth parameters (fresh weight, height, diameter) over time will be provided as output.
  • The climate control strategy will determine the use of resources, mainly energy (for heating, for electricity for artificial light) and therefore creates costs.
  • Fresh weight, height and diameter of the average lettuce plant are provided as the main output. These determine product price and therefore create income.
  • Teams will have to develop machine learning algorithms to feed the simple simulator with the optimised control parameters in order to maximise net profit.
  • Technical information on the working principle and an access key will be sent to all eligible teams at the start of the preparation phase.
  • During the preparation phase teams can interact with the simple simulator for algorithm development. During the Online Challenge this algorithm should be suitable to control the growth of a virtual crop in a virtual greenhouse under changed conditions (e.g. other weather conditions, different greenhouse type, different lettuce type) and limited time constraints.
  • There will be different versions of the simulators (A-D) with slightly different simulation parameters (e.g. other weather conditions, different greenhouse type, different lettuce type):
    • Simulator A will be avaible during the preparation phase of the Online Challenge form 1 June 0:00 GMT to 11 July 23:00 GMT, with limited access (typically 1000 times per day), to train the algorithms of the teams.
    • Simulator B will be available once every day from 12:00-13:00 GMT, to test the trained model, with limited access (typically 200 times per day). A public ranking board will be generated according to the net profit on simulator B. On July 11th 2021 23:00 GMT Simulator A and B will be closed.
    • On July 12th 00:00 GMT, Simulator C will be available for a period of 24 h, with limited access (typically 1000 times), to re-train the model.
    • On July 13th 12:00 GMT a private ranking board will be opend and simulator D will be provided to the teams with limited access (typically 200 times). Teams must submit the optimised control parameters in this new simulator version based on their developed algorithms in order to maximise net profit. After July 13th 13:00 the submitting will be rejected. Teams can only see their own scores in the private board before the announcement of the final results realised in simulator D on July 14th.
  • The teams with the highest net profit will be ranked first for this part of the Online Challenge. A total ranking will be made according to the private ranking board.
  • 1st Place - $ 4,000 and A wild-card for direct participation in the Growing Challenge
  • 2nd Place - $ 3,000
  • 3rd Place - $ 1,000