datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
imagenet_sketchImageNet-Sketch data set consists of 50000 images, 50 images for each of the 1000 ImageNet classes.
We construct the data set with Google Image queries "sketch of __", where __ is the standard class name.
We only search within the "black and white" color scheme. We initially query 100 images for every class,
and then manually clean the pulled images by deleting the irrelevant images and images that are for similar
but different classes. For some classes, there are less than 50 images after manually cleaning, and then we
augment the data set by flipping and rotating the images.sketch2stl-sketches-image
sketch2stl-sketches-image
36 original 224 x 224 line sketches, each labelled with whether the drawing is a flat outline meant to be
extruded or a depiction of a 3D object, plus 360 label-preserving augmented images.
1. Purpose
Built for Homework 1 of Designing with AI, in direct support of my Project 1 system: a Gradio app that turns
a 2D sketch into an editable, 3D-printable STL.
This dataset sits directly under that pipeline. The input to my system is a drawing… See the full description on the dataset page: https://huggingface.co/datasets/sunkaiwen/sketch2stl-sketches-image.sketch_val_new
sketch_val_new
Sketch-prompt validation: a red circle (which object) plus a blue arrow (where it
goes) drawn over a robot workspace, paired with a short, deliberately under-specified
caption ("put this there"). Each row is labelled valid / failure, and failures carry a
mode code from a shared 19-code taxonomy (taxonomy.json).
archive
rows
sketches
labels
human_275.zip
275
drawn by human participants
human-reviewed, 5 rounds
drawvla_8522_synthetic.zip
8,522
generated… See the full description on the dataset page: https://huggingface.co/datasets/skkyshine/sketch_val_new.shared-class-sketch-diverse
Shared-Class Sketch
Shared-Class Sketch is a deterministic synthetic corpus for testing whether a
model can represent sketch categories through reusable primitives and parts.
Every image is a 128 x 128 binary raster with white strokes on a black
background. No real sketch raster or stroke is copied into the dataset.
Classes and structural variation
The ten semantic classes are bicycle, truck, binoculars, face,
windmill, birthday_cake, butterfly, chandelier, flower… See the full description on the dataset page: https://huggingface.co/datasets/Rathgrith/shared-class-sketch-diverse.sketch2stl-datasets
sketch2stl datasets
CMU 24-679 Project 1 · Serena Sun & PK. Source CAD data: Fusion 360 Gallery Reconstruction r1.0.1 (Autodesk, license included).
Folder
For
Manual (counts toward the ≥500)
Synthetic / derived (training only)
ML1_stroke_recognizer/
ML 1: what kind of stroke is this? (line, arc, circle, rect, polyline) · trained from scratch
325 hand-drawn + 197 human-reviewed Fusion strokes = 522
29,387 synthetic strokes
ML2_smart_suggestions/
ML 2: what should… See the full description on the dataset page: https://huggingface.co/datasets/pakiino/sketch2stl-datasets.Step-by-Step_Sketch_Predictor_Dataset
Andrew Loomis Head Construction: Sequential Drawing Dataset
Overview
This dataset provides a comprehensive collection of 175 unique head drawing sequences based on the world-renowned Andrew Loomis construction method. Unlike static image datasets, this collection captures the temporal progression of a drawing—from initial geometric primitives to high-fidelity graphite renderings.
Purpose
Designed specifically for fine-tuning generative models… See the full description on the dataset page: https://huggingface.co/datasets/AqibRadius/Step-by-Step_Sketch_Predictor_Dataset.Experiment-Design-Sketch-Image-Classification-Dataset
Experiment Design Sketch Image Classification Dataset
In the field of industrial manufacturing, the design process often relies on a large number of design sketches that need to be quickly converted into actual engineering designs during the subsequent manufacturing stages. However, manually processing these sketches is often time-consuming and prone to errors, currently relying mainly on manual labeling and conversion by designers, which is inefficient and unstable. Existing… See the full description on the dataset page: https://huggingface.co/datasets/Mobiusi/Experiment-Design-Sketch-Image-Classification-Dataset.