Problem: The problem you are solving involves recognizing hand-drawn sketches using deep learning. Motivation: We chose this problem because visual communication is powerful, and enabling quick sketch recognition has broad applications.
Models: We experimented with a convolutional neural network (CNN) architecture. Specifically, you preprocess the sketches, extract contours, and use them alongside the original sketch as input to the CNN. The model is trained to predict the category labels.
Challenges: Challenges include handling abstract and sparse hand-drawn sketches, lack of visual clues, and variations in shape and abstraction.
Satisfaction: Achieving high recognition rates for hand-drawn sketches is a significant achievement.
Approach Recommendation: Start with a solid dataset, iterate on model architectures, and prioritize real-world usability.
Log in or sign up for Devpost to join the conversation.