Introduction
I Spy is a classic children's game where one player selects an object, mentions its color, and the opponent must guess what the object is. This project implements a digital version of the game using Python, OpenCV, and NumPy. It also leverages a pretrained COCO (Common Objects in Context) model capable of detecting and classifying everyday objects. The goal of the application is to allow a user to input a color and have the system intelligently guess the object they are thinking of based on that color.
Methods
The game begins with the user selecting an object in their environment and inputting its color via the console. Using OpenCV, the system converts the camera frame to HSV color space and applies a mask based on the chosen color. This isolates regions in the frame where that color appears.
To work around the challenge of mapping a color to a specific object, the application identifies bounding boxes around the masked areas. These bounding boxes represent segments of the frame where the specified color is found. The system then uses the COCO model to classify only the objects within these color-segmented bounding boxes.
The detected objects are stored in a list of strings, which the computer then uses to sequentially guess the user’s chosen item. The user responds with “yes” or “no” until the correct object is identified.
Conclusion
This application successfully integrates color-based masking and object classification to play a simple game of I Spy. It is capable of identifying and guessing objects of a specified color with a reasonable level of accuracy. In testing, the object recognition system achieved an accuracy rate of approximately 82%, making it effective for most common household or everyday objects.
While effective in its current form, the game could be further improved by implementing continuous learning mechanisms or fine-tuning the pretrained COCO model to enhance accuracy and responsiveness, particularly in more complex or ambiguous visual environments.
