Search for: All records

Schistosomiasis is a parasitic disease with significant global health and socio-economic implications. Drug discovery for schistosomiasis typically involves high-content whole-organism screening. In this approach, parasites are ex-posed to various chemical compounds and their systemic, whole-organism-level responses are captured via microscopy and analyzed to obtain a quanti-tative assessment of chemical effect. These effects are multidimensional and time-varying, impacting shape, appearance, and behavior. Accurate identifi-cation of object boundaries is essential for preparing images for subsequent analysis in high-content studies. Object segmentation is one of the most deeply studied problems in computer vision where recent efforts have incor-porated deep learning. Emerging results indicate that acquiring robust fea-tures in spectral domain using Fast Fourier Transform (FFT) within Deep Neural Networks (DNNs) can enhance segmentation accuracy. In this paper, we explore this direction further and propose a latent space Phase-Gating (PG) method that builds upon FFT and leverages phase information to effi-ciently identify globally significant features. While the importance of phase in analyzing signals has long been known, technical difficulties in calculat-ing phase in manners that are invariant to imaging parameters has limited its use. A key result of this paper is to show how phase information can be in-corporated in neural architectures that are compact. Experiments conducted on complex HCS datasets demonstrate how this idea leads to improved seg-mentation accuracy, while maintaining robustness against commonly en-countered noise (blurring) in HCS. The compactness of the proposed method also makes it well-suited for application specific architectures (ASIC) de-signed for high-content screening. 

In the last decade, convolutional neural networks (CNNs) have evolved to become the dominant models for various computer vision tasks, but they cannot be deployed in low-memory devices due to its high memory requirement and computational cost. One popular, straightforward approach to compressing CNNs is network slimming, which imposes an ℓ1 penalty on the channel-associated scaling factors in the batch normalization layers during training. In this way, channels with low scaling factors are identified to be insignificant and are pruned in the models. In this paper, we propose replacing the ℓ1 penalty with the ℓp and transformed ℓ1 (T ℓ1 ) penalties since these nonconvex penalties outperformed ℓ1 in yielding sparser satisfactory solutions in various compressed sensing problems. In our numerical experiments, we demonstrate network slimming with ℓp and T ℓ1 penalties on VGGNet and Densenet trained on CIFAR 10/100. The results demonstrate that the nonconvex penalties compress CNNs better than ℓ1 . In addition, T ℓ1 preserves the model accuracy after channel pruning, and ℓ1/2,3/4 yield compressed models with similar accuracies as ℓ1 after retraining. 

In this paper, we extend the traditional few-shot learning (FSL) problem to the situation when the source-domain data is not accessible but only high-level information in the form of class prototypes is available. This limited information setup for the FSL problem deserves much attention due to its implication of privacy-preserving inaccessibility to the source-domain data but it has rarely been addressed before. Because of limited training data, we propose a non-parametric approach to this FSL problem by assuming that all the class prototypes are structurally arranged on a manifold. Accordingly, we estimate the novel-class prototype locations by projecting the few-shot samples onto the average of the subspaces on which the surrounding classes lie. During classification, we again exploit the structural arrangement of the categories by inducing a Markov chain on the graph constructed with the class prototypes. This manifold distance obtained using the Markov chain is expected to produce better results compared to a traditional nearest- neighbor-based Euclidean distance. To evaluate our proposed framework, we have tested it on two image datasets – the large-scale ImageNet and the small-scale but fine-grained CUB-200. We have also studied parameter sensitivity to better understand our framework.