From 3bfb58448ad0e883597fcacee081589f2e68affb Mon Sep 17 00:00:00 2001 From: Juan Carlos Mier Date: Wed, 28 Apr 2021 13:43:13 -0700 Subject: [PATCH] Update markdown links format PiperOrigin-RevId: 370976587 --- CIQA/README.md | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/CIQA/README.md b/CIQA/README.md index 0d40249e9e6b..42b1c5d753ea 100644 --- a/CIQA/README.md +++ b/CIQA/README.md @@ -4,19 +4,19 @@ ## Overview The CIQA dataset is an open-sourced collection of labels to the popular -[Aesthetic Visual Analysis (AVA)] (https://paperswithcode.com/dataset/aesthetic-visual-analysis) +[Aesthetic Visual Analysis (AVA)](https://paperswithcode.com/dataset/aesthetic-visual-analysis) dataset. Images from AVA were sampled, compressed to different JPEG quality factors using the Tensorflow tf.image.adjust_jpeg_quality method and rated by human raters in a forced choice pairwise comparison study. This dataset is the result of the work in [Deep Perceptual Image Quality -Assessment for Compression] (https://arxiv.org/abs/2103.01114) +Assessment for Compression](https://arxiv.org/abs/2103.01114) ## AVA Image Sampling -The [Aesthetic Visual Analysis dataset (AVA)] (https://ieeexplore.ieee.org/document/6247954) +The [Aesthetic Visual Analysis dataset (AVA)](https://ieeexplore.ieee.org/document/6247954) is well suited for deep learning applied to aesthetic Image Quality Assessment (IQA) proven by its successful implementation in the -[NIMA] (https://arxiv.org/abs/1709.05424) model. +[NIMA](https://arxiv.org/abs/1709.05424) model. AVA contains ∼ 255,000 images rated based on aesthetic qualities by amateur photographers. Because these images are stored using JPEG compression, only the subset of images with a JPEG quality factor, Q, of 99 or more (near lossless)