Chapter 1. Introduction 1
5.2 Method
5.3.2 Descriptive statistic of features of image with annotations identified
image-based response. A list of features was provided in questionnaire consisting of the following items- 1) relevance, 2) uniqueness, 3) clarity, 4) sketching ability, 5) choice of colours, 6) language processing, and 7) narration (Al-Homoud, 2020; Charlet & Damnati, 2017; Chaudhuri et al., 2020, 2021b; Gagnon et al., 2019; Sangkloy et al., 2017; Sarkar &
Chakrabarti, 2011; Z. Wang et al., 2017; Xueqing et al., 2018). Initially, a pilot study was conducted to capture any other additional features to evaluate novelty from annotated image- based pattern of creative responses. The additional five features were as follows- 1) imaginative, 2) subject knowledge, 3) versatility, 4) presentation, and 5) refining. Imaginative indicate any annotated image-based pattern of creative responses representing creative and divergent ideas. Subject knowledge refers to adequate perception of domain of interest. A response that is adaptable based on requirement is considered versatile. Presentation is a feature in which a response is illustrated in a clear and concise manner. Refining indicates the iterative process by which a response attempts to improve gradually. Descriptive statistic of all features is illustrated in Table 5.1.
Table 5.1: Descriptive statistic of features for image with annotations Statistics
Relev ance
Unique ness
Clarit y
Sketchi ngAbili
ty
Choice ofColo
rs
Langu agePro
ces Nar rati on
Ima gina tive
Subject Knowl
edge Vers
atili ty
Pres enta tion
Refin ing
N
Valid 71 71 71 71 71 71 71 71 71 71 71 71
Missi
ng 0 0 0 0 0 0 0 0 0 0 0 0
Mean 1.00 1.04 4.73 4.75 4.72 1.04 4.68 4.21 4.20 4.73 4.76 4.66 Standard
Deviation 0 0.2 0.56 0.52 0.53 0.2 0.6 0.86 0.8 0.53 0.52 0.58 Median 1.00 1.00 5.00 5.00 5.00 1.00 5.00 4.00 4.00 5.00 5.00 5.00
Mode 1 1 5 5 5 1 5 5 5 5 5 5
Frequency of all twelve features in questionnaire captured from literature and identified from experts in a pilot study is illustrated in Table 5.2-5.13. It was captured using Likert-type scale with labels very important=1, slightly more important=2, important=3, slightly important=4, and not at all important=5. The summarized form of significance of features derived from descriptive statistics is illustrated in Figure 5.5. In the context of mass examination, most of the subjects selected relevance, uniqueness of a response, and language processing feature to evaluate novelty which is represented in dark green colour. Few subjects marked uniqueness of an annotated image-based pattern of creative responses as slightly more important. Other features such as clarity, sketching ability, choice of colours, narration, imaginative, subject knowledge, versatility, presentation, and refining were chosen as important and slightly important by a few of the subjects marked in light green and yellow, respectively; while most of them chose those as not at all important are marked in sky blue. Therefore, relevance, uniqueness, and language processing of an annotated image-based pattern of creative responses were considered inputs to the proposed model due to their relative higher frequency for evaluating novelty. The internal consistency of the questionnaire measured with Cronbach’s alpha was found to be 0.703 (Gil-Gómez et al., 2017).
Table 5.2: Frequency of relevance between question and annotated image-based pattern of creative responses
Relevance
Frequency Percent Valid Percent Cumulative Percent
Valid very important 71 100.0 100.0 100.0
Table 5.3: Frequency of uniqueness in annotated image-based pattern of creative responses Uniqueness
Frequency Percent Valid Percent Cumulative Percent Valid
very important 68 95.8 95.8 95.8
slightly more important 3 4.2 4.2 100.0
Total 71 100.0 100.0
Table 5.4: Frequency of clarity in annotated image-based pattern of creative responses Clarity
Frequency Percent Valid Percent Cumulative Percent
Valid
important 4 5.6 5.6 5.6
slightly important 11 15.5 15.5 21.1
not at all important 56 78.9 78.9 100.0
Total 71 100.0 100.0
Table 5.5: Frequency of sketching ability in annotated image-based pattern of creative responses
Sketching ability
Frequency Percent Valid Percent Cumulative Percent
Valid
important 3 4.2 4.2 4.2
slightly important 12 16.9 16.9 21.1
not at all important 56 78.9 78.9 100.0
Total 71 100.0 100.0
Table 5.6: Frequency of choice of colours in annotated image-based pattern of creative responses
Choice of colours
Frequency Percent Valid Percent Cumulative Percent
Valid
important 3 4.2 4.2 4.2
slightly important 14 19.7 19.7 23.9
not at all important 54 76.1 76.1 100.0
Total 71 100.0 100.0
Table 5.7: Frequency of language processing in annotated image-based pattern of creative responses
Language processing
Frequency Percent Valid Percent Cumulative Percent Valid
very important 68 95.8 95.8 95.8
slightly more important 3 4.2 4.2 100.0
Total 71 100.0 100.0
Table 5.8: Frequency of narration in annotated image-based pattern of creative responses Narration
Frequency Percent Valid Percent Cumulative Percent
Valid
important 5 7.0 7.0 7.0
slightly important 13 18.3 18.3 25.4
not at all important 53 74.6 74.6 100.0
Total 71 100.0 100.0
Table 5.9: Frequency ofimagination in annotated image-based pattern of creative responses Imagination
Frequency Percent Valid Percent Cumulative Percent
Valid
slightly more important 1 1.4 1.4 1.4
important 17 23.9 23.9 25.4
slightly important 19 26.8 26.8 52.1
not at all important 34 47.9 47.9 100.0
Total 71 100.0 100.0
Table 5.10: Frequency of subject knowledge in annotated image-based pattern of creative responses
Subject knowledge
Frequency Percent Valid Percent Cumulative Percent
Valid
slightly more important 1 1.4 1.4 1.4
important 14 19.7 19.7 21.1
slightly important 26 36.6 36.6 57.7
not at all important 30 42.3 42.3 100.0
Total 71 100.0 100.0
Table 5.11: Frequency of versatility in annotated image-based pattern of creative responses Versatility
Frequency Percent Valid Percent Cumulative Percent
Valid
important 3 4.2 4.2 4.2
slightly important 13 18.3 18.3 22.5
not at all important 55 77.5 77.5 100.0
Total 71 100.0 100.0
Table 5.12: Frequency of presentation in annotated image-based pattern of creative responses Presentation
Frequency Percent Valid Percent Cumulative Percent
Valid
important 3 4.2 4.2 4.2
slightly important 11 15.5 15.5 19.7
not at all important 57 80.3 80.3 100.0
Total 71 100.0 100.0
Table 5.13: Frequency of refining in annotated image-based pattern of creative responses Frequency Percent Valid Percent Cumulative Percent
Valid
important 4 5.6 5.6 5.6
slightly important 16 22.5 22.5 28.2
not at all important 51 71.8 71.8 100.0
Total 71 100.0 100.0
Figure 5.5: Summary of significance of features of image with annotations 0
10 20 30 40 50 60 70 80
Very important Slightly more important Important Slightly important Not at all important
Frequency
5.3.3 Result of prediction of image with annotations
The YOLO model (Gordon et al., 2018) was pre-trained on ImageNet dataset. Further, approximately 1,00,000 image with annotations were considered from MSCOCO dataset. After pre-processing, sketches were randomly distributed in a proportion of 80% and 20% for training and validation, respectively. The dataset comprised 80 object categories and 91 stuff categories. Each image was associated with five annotations. The model contained 24 convolution layers followed by two fully connected layers. The last layer i.e., the output layer calculated probabilities of belonging to a particular class and further computed coordinates of bounding boxes of objects present in an image. The prediction of this model is illustrated in Figure 5.6.
Figure 5.6: Prediction of annotated images with coordinate of bounding boxes