Reconstruction of kiwifruit fruit geometry using a CGAN trained on a synthetic dataset

Olatunji, Jamal R.; Redding, Gabe P.; Rowe, Candy; East, A.R.

doi:10.1016/j.compag.2020.105699

Cited by 24 publications

(6 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Prior to the start of the greenhouse crop experiment, teams could explore, build, and train algorithms using a virtual greenhouse environment emulated by the available climate-crop models [ 24 , 43 ]. The use of synthetic training datasets has been shown to be very useful in earlier applications, when real-world data are not quantitatively and qualitatively sufficient for training purposes [ 51 , 66 ]. In the real growing experiment, contextually relevant data was collected via standard sensors, next to that data was collected by teams with additional sensors of their preference (see Section 2.3 ) to improve and increase the efficiency and robustness of their algorithms.…”

Section: Discussionmentioning

confidence: 99%

Cherry Tomato Production in Intelligent Greenhouses—Sensors and AI for Control of Climate, Irrigation, Crop Yield, and Quality

Hemming

Zwart

Elings

et al. 2020

Sensors

View full text Add to dashboard Cite

Greenhouses and indoor farming systems play an important role in providing fresh and nutritious food for the growing global population. Farms are becoming larger and greenhouse growers need to make complex decisions to maximize production and minimize resource use while meeting market requirements. However, highly skilled labor is increasingly lacking in the greenhouse sector. Moreover, extreme events such as the COVID-19 pandemic, can make farms temporarily less accessible. This highlights the need for more autonomous and remote-control strategies for greenhouse production. This paper describes and analyzes the results of the second “Autonomous Greenhouse Challenge”. In this challenge, an experiment was conducted in six high-tech greenhouse compartments during a period of six months of cherry tomato growing. The primary goal of the greenhouse operation was to maximize net profit, by controlling the greenhouse climate and crop with AI techniques. Five international teams with backgrounds in AI and horticulture were challenged in a competition to operate their own compartment remotely. They developed intelligent algorithms and use sensor data to determine climate setpoints and crop management strategy. All AI supported teams outperformed a human-operated greenhouse that served as reference. From the results obtained by the teams and from the analysis of the different climate-crop strategies, it was possible to detect challenges and opportunities for the future implementation of remote-control systems in greenhouse production.

show abstract

Section: Discussionmentioning

confidence: 99%

Cherry Tomato Production in Intelligent Greenhouses—Sensors and AI for Control of Climate, Irrigation, Crop Yield, and Quality

Hemming

Zwart

Elings

et al. 2020

Sensors

View full text Add to dashboard Cite

show abstract

“…Fruit occlusions in canopy pose great challenges to accurate fruit detection/localization and crop yield estimation. Olatunji et al (2020) applied CGAN to reconstruct complete kiwifruit surfaces by translating occluded fruit into non-occluded images to address partial occlusion problems in fruit detection. By reconstructing missing surface information of occluded fruit, the fruit shape, surface area and weight could be predicted from the reconstructed images.…”

Section: Fruit Detectionmentioning

confidence: 99%

“…The authors however did not discuss the impact on model performance with different numbers of generated training images. Similar to Olatunji et al (2020), applied Pix2Pix (Isola et al, 2017) to translate occluded berry images due to leaves to non-occluded fruit images for estimating the number of berries, enabling better yield estimation.…”

Section: Fruit Detectionmentioning

confidence: 99%

“…applied Pix2Pix GAN to remove the leaves from the berry images to generate non-occluded berry images for accurate berry counting. Olatunji et al (2020) reconstructs the complete kiwifruit surface by translating occluded fruit into non-occluded images to address partial occlusion problems in fruit detection. More research on image processing like makeup removal can be beneficial to the application of GANs to visual recognition tasks like fruit detection and counting.…”

Section: Other Considerationsmentioning

confidence: 99%

See 1 more Smart Citation

Generative Adversarial Networks for Image Augmentation in Agriculture: A Systematic Review

Olaniyi¹,

Chen²,

Lu³

et al. 2022

Preprint

View full text Add to dashboard Cite

In agricultural image analysis, optimal model performance is keenly pursued for better fulfilling visual recognition tasks (e.g., image classification, segmentation, object detection and localization), in the presence of challenges with biological variability and unstructured environments. Large-scale, balanced and ground-truthed image datasets, however, are often difficult to obtain to fuel the development of advanced, high-performance models. As artificial intelligence through deep learning is impacting analysis and modeling of agricultural images, data augmentation plays a crucial role in boosting model performance while reducing manual efforts for data preparation, by algorithmically expanding training datasets. Beyond traditional data augmentation techniques, generative adversarial network (GAN) invented in 2014 in the computer vision community, provides a suite of novel approaches that can learn good data representations and generate highly realistic samples. Since 2017, there has been a growth of research into GANs for image augmentation or synthesis in agriculture for improved model performance. This paper presents an overview of the evolution of GAN architectures followed by a systematic review of their application to agriculture (https://github.com/Derekabc/GANs-Agriculture), involving various vision tasks for plant health, weeds, fruits, aquaculture, animal farming, plant phenotyping as well as postharvest inspection of fruits. Challenges and opportunities of GANs are discussed for future research.

show abstract

“…For the purposes of network training, a synthetic dataset is procedurally generated in Blender, in order to mitigate the cost of labeling a large training dataset [11], [12]. Increasingly popular synthetic datasets like [13], have recently found applications in agriculture for various crops and cultures [14], [15], [16], including a synthetic dataset for the C. annuum semantic segmentation tasks [17]. Transfer learning for the network first conducted on the synthetic dataset is followed by additional fine tuning on a small dataset of real, manually labeled images.…”

Section: A Fruit Detectionmentioning

confidence: 99%

Render-in-the-loop aerial robotics simulator: Case Study on Yield Estimation in Indoor Agriculture

Ivanović¹,

Polić²,

Tabak³

et al. 2022

Preprint

View full text Add to dashboard Cite

Inspired by recent promising results in sim-toreal transfer in deep learning we built a realistic simulation environment combining a Robot Operating System (ROS)compatible physics simulator (Gazebo) with Cycles, the realistic production rendering engine from Blender. The proposed simulator pipeline allows us to simulate near-realistic RGB-D images. To showcase the capabilities of the simulator pipeline we propose a case study that focuses on indoor robotic farming. We developed a solution for sweet pepper yield estimation task. Our approach to yield estimation starts with aerial robotics control and trajectory planning, combined with deep learningbased pepper detection, and a clustering approach for counting fruit. The results of this case study show that we can combine real time dynamic simulation with near realistic rendering capabilities to simulate complex robotic systems.

show abstract

Reconstruction of kiwifruit fruit geometry using a CGAN trained on a synthetic dataset

Cited by 24 publications

References 33 publications

Cherry Tomato Production in Intelligent Greenhouses—Sensors and AI for Control of Climate, Irrigation, Crop Yield, and Quality

Cherry Tomato Production in Intelligent Greenhouses—Sensors and AI for Control of Climate, Irrigation, Crop Yield, and Quality

Generative Adversarial Networks for Image Augmentation in Agriculture: A Systematic Review

Render-in-the-loop aerial robotics simulator: Case Study on Yield Estimation in Indoor Agriculture

Contact Info

Product

Resources

About