IngredSAM: Open-World Food Ingredient Segmentation via a Single Image Prompt DOI Creative Commons

Leyi Chen,

Bowen Wang, Jiaxin Zhang

et al.

Journal of Imaging, Journal Year: 2024, Volume and Issue: 10(12), P. 305 - 305

Published: Nov. 26, 2024

Food semantic segmentation is of great significance in the field computer vision and artificial intelligence, especially application food image analysis. Due to complexity variety food, it difficult effectively handle this task using supervised methods. Thus, we introduce IngredSAM, a novel approach for open-world ingredient segmentation, extending capabilities Segment Anything Model (SAM). Utilizing visual foundation models (VFMs) prompt engineering, IngredSAM leverages discriminative matchable features between single clean specific ingredients images guide generation accurate masks real-world scenarios. This method addresses challenges traditional dealing with diverse appearances class imbalances ingredients. Our framework demonstrates significant advancements without any training process, achieving 2.85% 6.01% better performance than previous state-of-the-art methods on both FoodSeg103 UECFoodPix datasets. exemplifies successful one-shot, paving way downstream applications such as enhancements nutritional analysis consumer dietary trend monitoring.

Language: Английский

Optimizing Deep Learning Models for Climate-Related Natural Disaster Detection from UAV Images and Remote Sensing Data DOI Creative Commons

Kim VanExel,

Samendra P. Sherchan, Siyan Liu

et al.

Journal of Imaging, Journal Year: 2025, Volume and Issue: 11(2), P. 32 - 32

Published: Jan. 24, 2025

This research study utilized artificial intelligence (AI) to detect natural disasters from aerial images. Flooding and desertification were two taken into consideration. The Climate Change Dataset was created by compiling various open-access data sources. dataset contains 6334 images UAV (unmanned vehicles) satellite then used train Deep Learning (DL) models identify disasters. Four different Machine (ML) used: convolutional neural network (CNN), DenseNet201, VGG16, ResNet50. These ML trained on our so that their performance could be compared. DenseNet201 chosen for optimization. All four performed well. ResNet50 achieved the highest testing accuracies of 99.37% 99.21%, respectively. project demonstrates potential AI address environmental challenges, such as climate change-related study’s approach is novel creating a new dataset, optimizing an model, cross-validating, presenting one DL detection. Three categories (Flooded, Desert, Neither). Our relates Environmental Sustainability. Drone emergency response would practical application project.

Language: Английский

Citations

0
Emergency Mapping for Flood Events Using Satellite Imagery (Optical and/or Synthetic Aperture Radar) and Benchmarking between Open Source and Commercial Geospatial Analysis Tools DOI Open Access
Nancy Alvan Romero,

Wilson A. Suarez Alayza

Published: July 23, 2024

Weather conditions appear to be undergoing significant deviations from the long-term average, marked by pronounced extremes of heat, prolonged droughts, and heightened rainfall occurring with greater frequency worldwide. Consequently, new patterns extreme weather are emerging, like unusual unorganized tropical cyclone called "Yaku", that influenced amount rain between March 6th 10th, 2023 hits more than 1000 districts in northwestern Peru. One district affected was Íllimo, Lambayeque province, due river pass through city, so “La Leche”, after continuous intensive overflow devastating his surrounding areas causing victims damage. This emergency provided excellent opportunity apply, on aforementioned areas, “change detection” technique, allows identifying change data obtained before events. In this research, optical (Sentinel-2, Landsat, MODIS Terra Aqua, PeruSAT-1) synthetic aperture radar (SAR) (Sentinel-1 COSMO SkyMed) were analyzed. A benchmark could also carried out open-source commercial spatial analysis tools. The results indicated an event obscured clouds do not allow their use, while SAR overcome clouds, can used for research. using Jaccard index score, Sentinel-1A set showed a 33% correlation, SkyMed demonstrated 38% match maps. sensitize population develop better management future events (floods) different

Language: Английский

Citations

1
Hybrid Deep Learning Model for Pancreatic Cancer Image Segmentation DOI
Wilson Bakasa, Clopas Kwenda, Serestina Viriri

et al.

Lecture notes in computer science, Journal Year: 2024, Volume and Issue: unknown, P. 14 - 24

Published: Oct. 2, 2024

Language: Английский

Citations

1
IngredSAM: Open-World Food Ingredient Segmentation via a Single Image Prompt DOI Creative Commons

Leyi Chen,

Bowen Wang, Jiaxin Zhang

et al.

Journal of Imaging, Journal Year: 2024, Volume and Issue: 10(12), P. 305 - 305

Published: Nov. 26, 2024

Food semantic segmentation is of great significance in the field computer vision and artificial intelligence, especially application food image analysis. Due to complexity variety food, it difficult effectively handle this task using supervised methods. Thus, we introduce IngredSAM, a novel approach for open-world ingredient segmentation, extending capabilities Segment Anything Model (SAM). Utilizing visual foundation models (VFMs) prompt engineering, IngredSAM leverages discriminative matchable features between single clean specific ingredients images guide generation accurate masks real-world scenarios. This method addresses challenges traditional dealing with diverse appearances class imbalances ingredients. Our framework demonstrates significant advancements without any training process, achieving 2.85% 6.01% better performance than previous state-of-the-art methods on both FoodSeg103 UECFoodPix datasets. exemplifies successful one-shot, paving way downstream applications such as enhancements nutritional analysis consumer dietary trend monitoring.

Language: Английский

Citations

0