Image-Label Recovery on Fashion Data Using Image Similarity from Triple Siamese Network
نویسندگان
چکیده
Weakly labeled data are inevitable in various research areas artificial intelligence (AI) where one has a modicum of knowledge about the complete dataset. One reasons for weakly AI is insufficient accurately data. Strict privacy control or accidental loss may also cause missing-data problems. However, supervised machine learning (ML) requires order to successfully solve problem. Data labeling difficult and time-consuming as it manual work, perfect results, sometimes human experts be involved (e.g., medical data). In contrast, unlabeled inexpensive easily available. Due there not being enough training data, researchers only obtain few points per category label. Training ML model from small set challenging task. The objective this recover missing labels dataset using state-of-the-art techniques semisupervised approach. novel convolutional neural network-based framework trained with instances class perform metric learning. then converted into graph signal, which recovered algorithm (RA) Fourier transform. proposed approach was evaluated on Fashion accuracy precision performed significantly better than networks other methods.
منابع مشابه
Image Classification using Transfer Learning from Siamese Networks based on Text Metadata Similarity
Convolutional neural networks learn about underlying image representations just by optimizing to a supervised classification. This project attempts to learn better features from images by training a network based on the similarity of pairs of images. Similarity of images will be computed based on the similarity of the text associated with the images as metadata, specifically captions for MSCOCO...
متن کاملA Novel Image Structural Similarity Index Considering Image Content Detectability Using Maximally Stable Extremal Region Descriptor
The image content detectability and image structure preservation are closely related concepts with undeniable role in image quality assessment. However, the most attention of image quality studies has been paid to image structure evaluation, few of them focused on image content detectability. Examining the image structure was firstly introduced and assessed in Structural SIMilarity (SSIM) measu...
متن کاملLearning Image Similarity Measures from Choice Data
We present a corpus of experimental data from psychometric studies on gamut mapping and demonstrate its use to develop image similarity measures. We investigate whether similarity measures based on luminance (SSIM) can be improved when features based on chroma and hue are added. Image similarity measures can be applied to automatically select a good image from a sample of transformed images.
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: Technologies (Basel)
سال: 2021
ISSN: ['2227-7080']
DOI: https://doi.org/10.3390/technologies9010010