Keynote Lectures

AI-Generated Content: Opportunities and Challenges
Abdenour Hadid, Sorbonne University of Abu Dhabi, United Arab Emirates, United Arab Emirates

AI and High Field MRI for Neurodegenerative Diseases
Christine Fernandez-Maloigne, University of Poitiers, France, France

RGB-Depth Imaging for High-Throughput Monitoring of Plant Growth
David Rousseau, Université D'angers, France, France

Abstract
It is becoming quite easy to generate realistic images and videos especially using diffusion-like models (e.g. DALL-E, GLIDE, Midjourney, Imagen, VideoPoet, Sora, Genie, etc.) due to their impressive generative capabilities. This creates a huge potential for a wide range of applications such as image editing, video production, content creation and digital marketing. Moreover, synthetically generated images and videos can be very useful for enhancing the training of AI models which usually require a large amount of data. However, these advances have also raised concerns about the potential misuse of these images and videos, including the creation of misleading content such as fake news and propaganda. So, one of the critical challenges associated with these advancements is the development of effective detection methods of synthetic images and videos. In the talk, we present the advances in automatically generating and editing images and videos, and discuss the limitations and challenges of such AI-generated content

Abstract
The exponential growth of computing resources has led to the potential improvement of diagnostic and therapeutic methods in medicine through the use of artificial intelligence (AI) methods. The field of imaging, particularly neuro MRI data, has been and will likely continue to be at the forefront of this revolution. These approaches, combined with new ultra-high field MRI techniques, will advance the understanding of brain biology by enabling virtual biopsy. This involves non-invasive sampling of the molecular environment with high spatial resolution, providing a better understanding of the underlying heterogeneous cellular and molecular processes. By providing in vivo markers of spatial and molecular heterogeneity, these AI-based tools have the potential to guide diagnoses and therapeutic pathways for neurodegenerative diseases more accurately for a patient and enable better dynamic treatment monitoring in the era of personalized medicine.

Abstract
Plants are complex 3D structures that present real challenges for computer vision, due to their self-similarity, the presence of self-occultation, or the low color contrasts between certain organs (particularly leaves). The need for plant imagery is growing for applications in biology and agriculture, where digital technology provides objective repeatability, high-throughput parallelization capability, observations on temporal-spatial scales and spectral ranges that are inaccessible to human eyes. In this context, the use of low-cost RGB-Depth cameras as introduced in 2012 [1], by hijacking the "Kinect" video game sensor at the time, is proving to be a very powerful tool for segmenting leaves, characterizing the 3D shape of plant cover or detecting the effect of biotic and abiotic stresses. This presentation summarizes the work [1-6] carried out by our group on this subject over more than a decade.

References:
[1] Chéné, Y., Rousseau, D., Lucidarme, P., Bertheloot, J., Caffier, V., Morel, P., ... & Chapeau-Blondeau, F. (2012). On the use of depth camera for 3D phenotyping of entire plants. Computers and Electronics in Agriculture, 82, 122-127.
[2] Chéné, Y., Belin, É., Rousseau, D., & Chapeau-Blondeau, F. (2013). Multiscale analysis of depth images from natural scenes: Scaling in the depth of the woods. Chaos, Solitons & Fractals, 54, 135-149.
[3] Chéné, Y., Rousseau, D., Belin, É., Garbez, M., Galopin, G., & Chapeau-Blondeau, F. (2016). Shape descriptors to characterize the shoot of entire plant from multiple side views of a motorized depth sensor. Machine Vision and Applications, 27, 447-461.
[4] Garbouge, H., Rasti, P., & Rousseau, D. (2021). Enhancing the Tracking of Seedling Growth Using RGB-Depth Fusion and Deep Learning. Sensors, 21(24), 8425.
[5] Couasnet, G., Cordier, M., Garbouge, H., Mercier, F., Pierre, D., El Ghaziri, A. Rousseau, D. (2023). Growth Data—An automatic solution for seedling growth analysis via RGB-Depth imaging sensors. SoftwareX, 24, 101572.
[6] Cordier M., Metuarea H., Bencheikh M., Rasti R., Torrez C., Rousseau D. (2023) Leaf segmentation of seedlings using foundation model on RGB-Depth images, ICCV Computer vision for plant phenotyping and agriculture workshop