Our daily lives depend on cereal crops such as wheat and barley, and our agricultural achievements depend on our ability to understand their phenotypic traits. These crops have awns, which are bristle-like extensions. The awns have multiple functions: protection, seed dispersal and photosynthesis. The edges have barbs, which are small hook-like structures on their surface. Although their importance is evident, analyzing these small structures has been challenging due to the lack of automated tools.
Accordingly, Plant Phenomics researchers have introduced BarbNet, a deep learning model specifically designed for automated detection and phenotyping of spikes in microscopic edge images. The researchers trained and validated the model using 348 diverse images representing various edge phenotypes with different spike sizes and densities. For the formulation of BarbNet, the researchers refined the U-net architecture, including modifications such as batch normalization, exclusion of dropout layers, increasing kernel size, and adjustments to model depth. These methodologies allow them to evaluate numerous characteristics, including size, shape, spine orientation, and additional characteristics such as glandular structures or pigment distribution.
Previously, scientists have used methods such as scanning electron microscopy to visualize edges. Although these techniques worked well, they could have been more efficient for high-throughput analysis. Plus, reviewing photos manually is time-consuming. So the researchers attempted to formulate a more sophisticated method for understanding the complicated inheritance patterns involved in the genetic basis of quill development.
The researchers evaluated the model on several benchmarks and found that while BarbNet demonstrated a 90% accuracy rate in detecting various edge phenotypes, it still has challenges in detecting small spikes and distinguishing densely packed ones. To overcome these obstacles and increase the accuracy and adaptability of awn analysis, the research team suggests expanding the training set and investigating different convolutional neural network (CNN) models. The researchers used binary cross-entropy loss and dice coefficient (DC) to train and validate the model. They found that it achieved a validation of 0.91 after 75 epochs.
Furthermore, they conducted a comparative study between automated segmentation results and manual real data, and the results show that BarbNet has a high agreement of 86% between BarbNet predictions and manual annotations. The researchers also investigated the classification of awn phenotypes based on genotype, focusing on four main awn phenotypes associated with two genes that regulate barb size and density.
In conclusion, BarbNet can be an important step in crop research as it offers powerful tools for automated edge analysis. By combining advanced deep learning techniques with genetic and phenotypic research, scientists can address the complexities of spike formation in cereal crops. BarbNet enables fast and accurate characterizations of edge and barb properties, promoting faster discoveries and improved breeding programs for higher yields.
Review the Paper. All credit for this research goes to the researchers of this project. Also, don't forget to join. our 33k+ ML SubReddit, 41k+ Facebook community, Discord channel, LinkedIn group, and Electronic newsletterwhere we share the latest news on ai research, interesting ai projects and more.
If you like our work, you'll love our newsletter.
Rachit Ranjan is a consulting intern at MarktechPost. He is currently pursuing his B.tech from the Indian Institute of technology (IIT), Patna. He is actively shaping his career in the field of artificial intelligence and data science and is passionate and dedicated to exploring these fields.
<!– ai CONTENT END 2 –>
