Sökning: "Label smoothing"

Hittade 3 uppsatser innehållade orden Label smoothing.

  1. 1. The effect of model calibration on noisy label detection

    Master-uppsats, KTH/Skolan för elektroteknik och datavetenskap (EECS)

    Författare :Max Joel Söderberg; [2023]
    Nyckelord :Area Under the Margin; Calibration; Image classification; Label smoothing; Noisy labels; Overconfidence; Area under marginalen; Kalibrering; Bildklassificering; Etikettutjämning; Brusiga etiketter; Övertro;

    Sammanfattning : The advances in deep neural networks in recent years have opened up the possibility of using image classification as a valuable tool in various areas, such as medical diagnosis from x-ray images. However, training deep neural networks requires large amounts of annotated data which has to be labelled manually, by a person. LÄS MER

  3. 2. Writer identification using semi-supervised GAN and LSR method on offline block characters

    Uppsats för yrkesexamina på avancerad nivå, Högskolan i Halmstad/Akademin för informationsteknologi

    Författare :Adrian Hagström; Rustam Stanikzai; [2020]
    Nyckelord :Computer vision; Deep learning; GAN; Generative adveserial network; Semi-supervised learning; LSR; Machine learning; ResNet; AI; Writer authentication; Writer identification; Datorseende; Djup lärning; GAN; Maskininlärning; AI; Skrvning autentisering; Skrvning identifiering;

    Sammanfattning : Block characters are often used when filling out forms, for example when writing ones personal number. The question of whether or not there is recoverable, biometric (identity related) information within individual digits of hand written personal numbers is then relevant. LÄS MER

  4. 3. Automatic Lane Labeling Using an Off-line Smoothing Stereo-lane Algorithm

    Master-uppsats, Linköpings universitet/Reglerteknik

    Författare :Mikael Hartman; [2017]
    Nyckelord :;

    Sammanfattning : The goal of this thesis has been to create a map of road lanes from piece-wisedetections. Using images from a stereo camera, an existing algorithm is able todetect lane tracks in real-time. With a basis spline (B-spline) representation ofthe tracks, these curve features are converted into point features. LÄS MER