More Like this

1. Structure-Aware Image Segmentation with Homotopy Warping

   Hu, Xiaoling (December 2022, The Thirty-sixth Conference on Neural Information Processing Systems)

   Besides per-pixel accuracy, topological correctness is also crucial for the segmentation of images with fine-scale structures, e.g., satellite images and biomedical images. In this paper, by leveraging the theory of digital topology, we identify pixels in an image that are critical for topology. By focusing on these critical pixels, we propose a new homotopy warping loss to train deep image segmentation networks for better topological accuracy. To efficiently identify these topologically critical pixels, we propose a new algorithm exploiting the distance transform. The proposed algorithm, as well as the loss function, naturally generalize to different topological structures in both 2D and 3D settings. The proposed loss function helps deep nets achieve better performance in terms of topology-aware metrics, outperforming state-of-the-art structure/topology-aware segmentation methods. 

   more » « less
2. Growth chamber reciprocal transplant dataset

   https://doi.org/10.5061/dryad.4mw6m90kb  

   Garrison, Ava (October 2024, Dryad)

   Phenotypic plasticity can alter traits that are crucial to population\n establishment in a new environment, before adaptation can occur. How often\n phenotypic plasticity enables subsequent adaptive evolution is unknown,\n and examples of the phenomenon are limited. We investigated the hypothesis\n of plasticity-mediated persistence as a means of colonization of\n agricultural fields in one of the world’s worst weeds, Raphanus\n raphanistrum ssp. raphanistrum. Using non-weedy native populations of the\n same species and subspecies as a comparison, we tested for\n plasticity-mediated persistence in a growth chamber reciprocal transplant\n experiment. We identified traits with genetic differentiation between the\n weedy and native ecotypes as well as phenotypic plasticity between growth\n chamber environments. We found that most traits were both plastic and\n differentiated between ecotypes, with the majority plastic and\n differentiated in the same direction. This suggests that phenotypic\n plasticity may have enabled radish populations to colonize and then adapt\n to novel agricultural environments."],"TechnicalInfo":["# Growth Chamber Reciprocal Transplant Dataset\n [https://doi.org/10.5061/dryad.4mw6m90kb](https://doi.org/10.5061/dryad.4mw6m90kb) This dataset contains the phenotypic data collected from plants grown in the growth chamber reciprocal transplant experiment, as well as the conditions in the growth chambers. ## Description of the data and file structure The dataset contains three sheets: "Chamber Conditions", "Main Data", and "Leaf Data" (although much of the information in "Leaf Data" has been incorporated into "Main data") ### Chamber Conditions This sheet contains the temperature and day length set points for each chamber each week. All temperature and day length information from the two weather stations used (LGER and KBTL) were collected from [www.wunderground.com](http://www.wunderground.com). Variables: * Granada, Spain (LGER) - the dates from which we collected temperature and day length information from the Grenada, Spain weather station (LGER) to simulate in the Winter Annual chamber * HighTempGrenada - the Winter Annual chamber's daytime set points, based on the average maximum temperature in Grenada on a given day * LowTempGrenada - the Winter Annual chamber's nighttime set points, based on the average minimum temperature in Grenada on a given day * DayLengthGrenada - the length of time the Winter Annual chamber was in its day cycle (lights on and typically higher temps), based on length of visible light in Grenada * DayStartGrenada - programed start of day time in the Winter Annual growth chamber * DayEndGrenada - programmed end of day time in the Winter Annual growth chamber * Date Set - the real-life date on which we changed the chamber conditions. * Augusta, MI (KBTL) - the dates from which we collected temperature and day length information from the Augusta, MI, USA weather station (KBTL) to simulate in the Spring Annual chamber * HighTempAugusta - the Spring Annual chamber's daytime set points, based on the average maximum temperature in Augusta on a given day * LowTempAugusta - the Spring Annual chamber's nighttime set points, based on the average minimum temperature in Augusta on a given day * DayLengthAugusta - the length of time the Spring Annual chamber was in its day cycle (lights on and typically higher temps), based on length of visible light in Augusta * DayStartAugusta - programed start of day time in the Spring Annual growth chamber * DayEndAugusta - programmed end of day time in the Spring Annual growth chamber ### Main Data This sheet contains all of the data used in our analyses, as well as descriptors for plants and growth chambers. Variables: * Chamber # - the number designation of the four growth chambers used in this study * Environment - the growing conditions in a given growth chamber, with "Winter Annual" corresponding to the "Grenada, Spain (LGER)" columns in Chamber Conditions, and "Spring Annual" corresponding to "Augusta, MI (KBTL)" * Ecotype - variety of* R. raphanistrum*, either weedy or native * Population - the six source populations used in this study identified by their location codes, with the final two letters denoting country or state (FR=France, ES=Spain, NY=New York, NC=North Carolina) and the first two letters denoting a specific location in those areas (available in Table 1 of the manuscript) * Matriline -  a line number is listed when discrete matrilines are known from field collections, but not for seeds collected in bulk (in which case the cell will be blank) * Flat - plants were arranged into four flats in each chamber, and the flats within a chamber were each assigned a number (1-4) * Position - the position of each plant within a flat was also tracked and pots were assigned a position number (1-35) * Pot# - Number assigned to each plant to give it a unique identifier -- for plants with individual matrilines tracked, pot # only went up to 2, while plants with unknown matrilines had pot numbers up to 40 to ensure individuals could be tracked * Plant Date - the date seeds were sown into each pot * Germ[1-5] - the date that each one of 5 seeds planted emerged as a germinant -- blank cells indicate that a germinant did not emerge * Plant Kept - the emergence date of the single plant that remained in the pot after excess germinants were thinned; missing values mean no germinants emerged or did not survive past the seedling stage * Days to Emergence - calculated as the day of emergence minus the planting date; missing values mean no germinants emerged or did not survive past the seedling stage * Rosette Photo Date - the date on which overhead and side photos of plants were taken, also the day the plants first showed signs of bolting (buds visible); missing values mean the plant did not survive to bolting * \\# Rosette Leaves - the number of leaves in the basal rosette, counted on the day of bolting; missing values mean the plant did not survive to bolting * Rosette Height - the vertical height of the tallest free-standing basal rosette leaf, measured from the height of the soil (cm); missing values mean the plant did not survive to bolting * 1st flower date - the date on which the first flower on a plant opened; missing values mean the plant did not survive to flowering * Days to First Flower - calculated as 1st flower date minus emergence date; missing values mean the plant did not survive to flowering * 1st Flower Height - measured on the first flower date, it is the vertical distance from the soil to the point at which the first open flower's pedicel connects to the main stalk (cm); missing values mean the plant did not survive to flowering * Ovule # - collected from typically the third flower to open, it is the number of ovules in one flower of a given plant; missing values mean the plant did not survive to flowering or ovules were not clearly visible * Notes - any additional information on a plant that we tracked * Blossom Photo Date - the date on which we took top and side photographs of at least the third flower to open, taken at the same time that ovule number was counted; missing values mean the plant did not survive to flowering  * PetalLength - measured using a top-view photo in Image J, the distance from the tip of the petal to where it meets the floral tube in the center of the floral display (mm); missing values mean the plant did not survive to flowering or the view in the photo was obscured so the measurement could not be taken * PetalWidth - measured using a top-view photo in Image J, the distance from the widest part of the petal, perpendicular to the line measured for petal length (mm); missing values mean the plant did not survive to flowering or the view in the photo was obscured so the measurement could not be taken * Tube - measured using a side-view photo in Image J, the length of the most clearly visible petal from where it meets the pedicel to the apex of its curve outward (mm); missing values mean the plant did not survive to flowering or the view in the photo was obscured so the measurement could not be taken * LAnther - measured using a side-view photo in Image J, the length of the anther of the long stamen from where it meets its filament to its tip (mm); missing values mean the plant did not survive to flowering or the view in the photo was obscured so the measurement could not be taken * LFilament - measured using a side-view photo in Image J, the length of the frontmost (closest to the camera) long filament from where it meets the pedicel to where it meets its anther (mm); missing values mean the plant did not survive to flowering or the view in the photo was obscured so the measurement could not be taken * SAnther - measured using a side-view photo in Image J, the length of the anther of the short stamen from where it meets its filament to its tip (mm); missing values mean the plant did not survive to flowering or the view in the photo was obscured so the measurement could not be taken * SFilament - measured using a side-view photo in Image J, the length of the frontmost (closest to the camera) short filament from where it meets the pedicel to where it meets its anther (mm); missing values mean the plant did not survive to flowering or the view in the photo was obscured so the measurement could not be taken * Pistil - measured using a side-view photo in Image J, the length of the pistil made by drawing a line down the center of the pistil from the top of the stigma to where it meets the pedicel (mm); missing values mean the plant did not survive to flowering or the view in the photo was obscured so the measurement could not be taken * AntherExsertion - calculated as long filament length minus the tube length (mm); missing values mean the plant did not survive to flowering or that either one of the values needed for the measurement was missing * AntherSeparation - calculated as long filament length minus the short filament length (mm); missing values mean the plant did not survive to flowering or one or that either one of the values needed for the measurement was missing * FlowerSize - the geometric mean of all floral traits (excluding anther exsertion and anther separation; mm); missing values mean the plant did not survive to flowering or one or more flower trait was missing * LeafWidth - measured using either a top-view or side-view photo in Image J, the distance between each edge of the leaf measured at its widest point, with the line being perpendicular to the central leaf vein on the largest fully visible leaf (more information in the Leaf Data sheet; cm); missing values mean the plant did not survive to bolting or a picture was not taken * LeafLength -  measured using either a top-view or side-view photo in Image J using the segmented line tool, follow the central vein of the largest visible leaf from the center of the rosette to the tip of the leaf (more information in the Leaf Data sheet; cm); missing values mean the plant did not survive to bolting or a picture was not taken ### Leaf Data This sheet includes some additional information about Leaf Length and Leaf Width measurements. Side image was only used when leaf was not flat or clearly visible in the top image. Variables: * Top Photo Image - image ID of the top view photo of the plant being measured * Ecotype - the ecotype of the plant (more information in Main Data) * Population - the population that the plant belongs to (more information in Main Data) * Plant Label - the label visible in the image -- includes population, matriline (when available), and pot # * Leaf Width (cm) - measured using the top-view photo in Image J, the distance between each edge of the leaf measured at its widest point, with the line being perpendicular to the central leaf vein on the largest fully visible leaf; missing values mean that a picture was not taken or the leaf was obscured in the top view photo * Leaf Length 1 (cm) - measured using the top-view photo in Image J using the segmented line tool, follow the central vein of the largest visible leaf from the center of the rosette to the tip of the leaf; missing values mean that a picture was not taken or the leaf was obscured in the top view photo * Side Photo Image - Image ID of the side view photo of the plant being measured; side image was only used when leaf was not flat or clearly visible in the top image, so missing values indicate that the length and width of the leaf could be reliably measured using the top view photo * Leaf Length 2 (cm) - measured using the side-view photo in Image J using the segmented line tool, follow the central vein of the largest visible leaf from the center of the rosette to the tip of the leaf; missing values mean that a picture was not taken or that the length of the leaf could be reliably measured using the top view photo * Leaf Width 2 (cm) - measured using the side-view photo in Image J, the distance between each edge of the leaf measured at its widest point, with the line being perpendicular to the central leaf vein on the largest fully visible leaf; missing values mean that a picture was not taken or that the width of the leaf could be reliably measured using the top view photo * Notes - any additional information about the the measurement of a particular plants' leaf length or width"]} 

   more » « less
3. Weakly-Supervised Object Representation Learning for Few-Shot Semantic Segmentation

   Ying, Xiaowen; Li, Xin; Chuah, Mooi Choo (January 2021, IEEE Winter Conference on Applications of Computer Vision)

   null (Ed.)

   Training a semantic segmentation model requires large densely-annotated image datasets that are costly to obtain. Once the training is done, it is also difficult to add new ob- ject categories to such segmentation models. In this pa- per, we tackle the few-shot semantic segmentation prob- lem, which aims to perform image segmentation task on un- seen object categories merely based on one or a few sup- port example(s). The key to solving this few-shot segmen- tation problem lies in effectively utilizing object informa- tion from support examples to separate target objects from the background in a query image. While existing meth- ods typically generate object-level representations by av- eraging local features in support images, we demonstrate that such object representations are typically noisy and less distinguishing. To solve this problem, we design an ob- ject representation generator (ORG) module which can ef- fectively aggregate local object features from support im- age(s) and produce better object-level representation. The ORG module can be embedded into the network and trained end-to-end in a weakly-supervised fashion without extra hu- man annotation. We incorporate this design into a modified encoder-decoder network to present a powerful and efficient framework for few-shot semantic segmentation. Experimen- tal results on the Pascal-VOC and MS-COCO datasets show that our approach achieves better performance compared to existing methods under both one-shot and five-shot settings. 

   more » « less
4. An Efficient Shift-and-stack Algorithm Applied to Detection Catalogs

   https://doi.org/10.3847/1538-3881/ae0e1a  

   Stetzler, Steven; Jurić, Mario; Bernardinelli, Pedro H; Bektešević, Dino; Chandler, Colin Orion; Connolly, Andrew J; Adams, Fred C; Fuentes, Cesar; Gerdes, David W; Holman, Matthew J; et al (November 2025, The Astronomical Journal)

   Abstract The boundary of solar system object discovery lies in detecting its faintest members. However, their discovery in detection catalogs from imaging surveys is fundamentally limited by the practice of thresholding detections at signal-to-noise (SNR) ≥ 5 to maintain catalog purity. Faint moving objects can be recovered from survey images using the shift-and-stack algorithm, which coadds pixels from multi-epoch images along a candidate trajectory. Trajectories matching real objects accumulate signal coherently, enabling high-confidence detections of very faint moving objects. Applying shift-and-stack comes with high computational cost, which scales with target object velocity, typically limiting its use to searches for slow-moving objects in the outer solar system. This work introduces a modified shift-and-stack algorithm that trades sensitivity for speedup. Our algorithm stacks low-SNR detection catalogs instead of pixels, the sparsity of which enables approximations that reduce the number of stacks required. Our algorithm achieves real-world speedups of 10–10³× over image-based shift-and-stack while retaining the ability to find faint objects. We validate its performance by recovering synthetic inner and outer solar system objects injected into images from the DECam Ecliptic Exploration Project. Exploring the sensitivity–compute time trade-off of this algorithm, we find that our method achieves a speedup of ∼30× with 88% of the memory usage while sacrificing 0.25 mag in depth compared to image-based shift-and-stack. These speedups enable the broad application of shift-and-stack to large-scale imaging surveys and searches for faint inner solar system objects. We provide a reference implementation via thefind-asteroidsPython package and this URL:https://github.com/stevenstetzler/find-asteroids. 

   more » « less
5. A Continual Learning Framework for Uncertainty-Aware Interactive Image Segmentation

   Zheng, Ervine; Yu, Qi; Li, Rui; Shi, Pengcheng; Haake, Anne (May 2021, Proceedings of the AAAI Conference on Artificial Intelligence)

   null (Ed.)

   Deep learning models have achieved state-of-the-art performance in semantic image segmentation, but the results provided by fully automatic algorithms are not always guaranteed satisfactory to users. Interactive segmentation offers a solution by accepting user annotations on selective areas of the images to refine the segmentation results. However, most existing models only focus on correcting the current image’s misclassified pixels, with no knowledge carried over to other images. In this work, we formulate interactive image segmentation as a continual learning problem and propose a framework to effectively learn from user annotations, aiming to improve the segmentation on both the current image and unseen images in future tasks while avoiding deteriorated performance on previously-seen images. It employs a probabilistic mask to control the neural network’s kernel activation and extract the most suitable features for segmenting images in each task. We also design a task-aware architecture to automatically infer the optimal kernel activation for initial segmentation and subsequent refinement. Interactions with users are guided through multi-source uncertainty estimation so that users can focus on the most important areas to minimize the overall manual annotation effort. Extensive experiments are performed on both medical and natural image datasets to illustrate the proposed framework’s effectiveness on basic segmentation performance, forward knowledge transfer, and backward knowledge transfer. 

   more » « less