Aesthetic Image Captioning From Weakly-Labelled Photographs

TL;DR

This paper introduces a new large-scale dataset for aesthetic image captioning created through an automatic cleaning process, and proposes a weakly supervised method to learn aesthetic features without needing detailed annotations.

Contribution

It presents a novel dataset, AVA-Captions, and a weakly supervised training strategy for aesthetic feature extraction in image captioning.

Findings

The dataset contains 230,000 images with 5 captions each.

The weakly supervised method effectively learns aesthetic representations.

Automatic metrics and subjective evaluations validate the approach.

Abstract

Aesthetic image captioning (AIC) refers to the multi-modal task of generating critical textual feedbacks for photographs. While in natural image captioning (NIC), deep models are trained in an end-to-end manner using large curated datasets such as MS-COCO, no such large-scale, clean dataset exists for AIC. Towards this goal, we propose an automatic cleaning strategy to create a benchmarking AIC dataset, by exploiting the images and noisy comments easily available from photography websites. We propose a probabilistic caption-filtering method for cleaning the noisy web-data, and compile a large-scale, clean dataset "AVA-Captions", (230, 000 images with 5 captions per image). Additionally, by exploiting the latent associations between aesthetic attributes, we propose a strategy for training the convolutional neural network (CNN) based visual feature extractor, the first component of the AIC framework. The strategy is weakly supervised and can be effectively used to learn rich aesthetic representations, without requiring expensive ground-truth annotations. We finally show-case a thorough analysis of the proposed contributions using automatic metrics and subjective evaluations.