Search Constraints

Start Over
Filtering by: Creator Krankina, Olga N. Remove constraint Creator: Krankina, Olga N. Committee Member Li, Fuxin Remove constraint Committee Member: Li, Fuxin

Search Results

  • Pixel- and Frame-level Video Labeling using Spatial and Temporal Convolutional Networks

    Creator:
    Lei, Peng
    Abstract:
    This dissertation addresses the problem of video labeling at both the frame and pixel levels using deep learning. For pixel-level video labeling, we have studied two problems: i) Spatiotemporal video segmentation and ii) Boundary detection and boundary flow estimation. For the problem of spatiotemporal video segmentation, we have developed recurrent...
    Resource Type:
    Dissertation
    Full Text:
    lengths of T/2 n and T , which are computed by the previous layer L− 1. In (a), the output features are

  • Action Segmentation with Limited Supervision

    Creator:
    Li, Jun
    Abstract:
    In this dissertation, we address action segmentation in videos under limited supervision. The goal of action segmentation is to predict an action class for each frame of a video. The limited supervision means ground truth labels of video frames are not available in training. We focus on three types of...
    Resource Type:
    Dissertation
    Full Text:
    for computing the cross-entropy loss and n-pair loss. . . . . . . . . . . . . . . . . 37 4.9 Our SCV

  • A Deep Action Segmentation and Its Explanation with a Dictionary of Meaningful Attention Maps

    Creator:
    Trigkakis, Dimitrios
    Abstract:
    This thesis addresses the problem of temporal action segmentation in videos, where the goal is to label every video frame with the appropriate action class present. We focus on the domain of NFL football videos, where action classes represent common football play types. For action segmentation, we use a temporal...
    Resource Type:
    Masters Thesis
    Full Text:
    VAE Loss = N∑ i=1 li li(θ, φ) = −Ez∼qθ(z|xi)[log pφ(xi|z)] +KL(qθ(z|xi)||p(z)) Our intent is to

  • Robust and Efficient Classification of Videos in the Wild

    Creator:
    Mahasseni, Behrooz
    Abstract:
    Recognizing human actions in videos is a long-standing problem in computer vision with a wide range of applications including video surveillance, content retrieval, and sports analysis. This thesis focuses on addressing efficiency and robustness of video classification in unconstrained real-world settings. The thesis work can be broadly divided into four...
    Resource Type:
    Dissertation
    Full Text:
    action occurs in the video, it occupies only a subset of K � N windows, A = {Vk : Vk ∈ V, k = 1, ...,K

  • The Familiarity Hypothesis: Explaining the Behavior of Deep Learning Methods in Out-of-Distribution Detection

    Creator:
    Guyer, Alexander
    Abstract:
    Many object recognition applications require detecting and responding to objects drawn from a different distribution from that of the training data. This task is referred to as out-of-distribution (OOD) detection, and it is often formulated as an outlier detection problem wherein the probability distribution of the known data P(X) is...
    Resource Type:
    Masters Thesis
    Full Text:
    plotting UMAP visualizations of the 9 0 50 100 150 200 Epochs 0 2 4 6 8 10 M ea n Ve cto r N

  • Fine-Grained Object Recognition Under Limited Training Data

    Creator:
    Lam, Michael Q.
    Abstract:
    This dissertation addresses object recognition in challenging settings, where distinct object classes are visually very similar (e.g., species of birds and insects) and/or access to training examples of object classes is limited (e.g., due to the associated high costs of data annotation). In this dissertation, we present a variety of...
    Resource Type:
    Dissertation
    Full Text:
    components N . . . . . . . . . . . . . . 88 6.5 Qualitative results on Human3.6M

  • Part-based and Uncertainty-Aware Few-shot Object Segmentation in Images

    Creator:
    Nguyen, Khoi D.
    Abstract:
    This dissertation addresses few-shot object segmentation in images. The goal of segmentation is to label every image pixel with a class of the object occupying that pixel, where the class may represent a semantic object category or instance. In few-shot segmentation, training and test datasets have different classes. Every new...
    Resource Type:
    Dissertation
    Full Text:
    number of experts N in the one-shot setting on PASCAL-5i. Part 0 – Part 3 denote the four folds of cross

  • Semantic image segmentation using domain constraints

    Creator:
    Roy, Anirban
    Abstract:
    This dissertation addresses the problem of semantic labeling of image pixels. In the course of our work, we considered different types of semantic labels, including object classes (e.g., car, person), 3D depth values (in the range 0 to 80 meters), and affordance classes (e.g., walkable, sittable). Semantic pixel labeling is...
    Resource Type:
    Dissertation
    Full Text:
    organized in a graph, G = (V,E). V is a set of nodes, i = 1, . . . , n, |V | = n, corresponding to