1
|
AI Ekphrasis: Multi-Modal Learning with Foundation Models for Fine-Grained Poetry Retrieval. ELECTRONICS 2022. [DOI: 10.3390/electronics11081275] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Track Full Text] [Subscribe] [Scholar Register] [Indexed: 11/16/2022]
Abstract
Artificial intelligence research in natural language processing in the context of poetry struggles with the recognition of holistic content such as poetic symbolism, metaphor, and other fine-grained attributes. Given these challenges, multi-modal image–poetry reasoning and retrieval remain largely unexplored. Our recent accessibility study indicates that poetry is an effective medium to convey visual artwork attributes for improved artwork appreciation of people with visual impairments. We, therefore, introduce a deep learning approach for the automatic retrieval of poetry suitable to the input images. The recent state-of-the-art CLIP provides a way for multi-modal visual and text features matched using cosine similarity. However, it lacks shared cross-modality attention features to model fine-grained relationships. The proposed approach in this work takes advantage of strong pre-training of the CLIP model and overcomes its limitations by introducing shared attention parameters to better model the fine-grained relationship between both modalities. We test and compare our proposed approach using the expertly annotated MiltiM-Poem dataset, which is considered the largest public image–poetry pair dataset for English poetry. The proposed approach aims to solve the problems of image-based attribute recognition and automatic retrieval for fine-grained poetic verses. The test results reflect that the shared attention parameters alleviate fine-grained attribute recognition, and the proposed approach is a significant step towards automatic multi-modal retrieval for improved artwork appreciation of people with visual impairments.
Collapse
|
2
|
Abstract
The development of assistive technologies is improving the independent access of blind and visually impaired people to visual artworks through non-visual channels. Current single modality tactile and auditory approaches to communicate color contents must compromise between conveying a broad color palette, ease of learning, and suffer from limited expressiveness. In this work, we propose a multi-sensory color code system that uses sound and scent to represent colors. Melodies express each color’s hue and scents the saturated, light, and dark color dimensions for each hue. In collaboration with eighteen participants, we evaluated the color identification rate achieved when using the multi-sensory approach. Seven (39%) of the participants improved their identification rate, five (28%) remained the same, and six (33%) performed worse when compared to an audio-only color code alternative. The participants then evaluated and compared a color content exploration prototype that uses the proposed color code with a tactile graphic equivalent using the System Usability Scale. For a visual artwork color exploration task, the multi-sensory color code integrated prototype received a score of 78.61, while the tactile graphics equivalent received 61.53. User feedback indicates that the multi-sensory color code system improved the convenience and confidence of the participants.
Collapse
|
3
|
Abstract
Contemporary art is evolving beyond simply looking at works, and the development of various sensory technologies has had a great influence on culture and art. Accordingly, opportunities for the visually impaired to appreciate visual artworks through various senses such as auditory and tactile senses are expanding. However, insufficient sound expression and lack of portability make it less understandable and accessible. This paper attempts to convey a color and depth coding scheme to the visually impaired, based on alternative sensory modalities, such as hearing (by encoding the color and depth information with 3D sounds of audio description) and touch (to be used for interface-triggering information such as color and depth). The proposed color-coding scheme represents light, saturated, and dark colors for red, orange, yellow, yellow-green, green, blue-green, blue, and purple. The paper’s proposed system can be used for both mobile platforms and 2.5D (relief) models.
Collapse
|