Product ideation sketches

Published: 22 August 2022| Version 1 | DOI: 10.17632/ng42jh2pfb.1
Contributors:
Sunny Prajapati,
Tarun Kumar,
Rahul Bhaumik

Description

Folder ‘sketches’ containing 510 images of ideation sketches drawn during a design activity. All images are resized to 220 by 150 pixels with minimal white spaces, and no distortions in the original proportions. The folder ‘attributes’ contains sketches segregated based on their assigned attributes in respective folders. The sub-folders are named as per the following convention “attribute1_attribute2” (where attribute1 and attribute2 are the attributes assigned to the sketches, these folders are: a) cute_lively b) elegant_beautiful c) simple_cute d) stable_masculine e) strong_compact Post the analysis of the attribute texts, the probabilities in each emotion class (based on the Plutchik’s model) for the images are recorded with the attributes, in the file “bottle_emotions.csv”.

Files

Steps to reproduce

Steps to reproduce: 1. Hierarchical clustering with following parameters: a. Distance metric: cosine distances b. Distance measurement: ward linkage c. Cluster selection method: top 6 cluster 2. Attribute assignment to each cluster by creating a separate text feature containing attribute(s) unique to each cluster. The images belonging to each cluster can be saved separately. 3. Pre-processing the attribute features: a. Transformations: conversion to lowercase, removal of accents b. Splitting texts by words and keep punctuations c. Filtering the language specific stop words. In this case English language was used 4. Tweet profiler for detecting emotion states in the text a. Emotions: states based on Plutchik’s model b. Output as most probable emotion for the attributes, or probabilities for each category

Institutions

Indian Institute of Science, PES University

Categories

Computer-Aided Design, Machine Learning, Aesthetics in Design, Convolutional Neural Network

License