Skip to content

HHIR/CV_proj

Repository files navigation

Evaluating object detection with custom YOLOv8 architecture for enhanced speed and accuracy

This project, completed as part of the Computer Vision course at Sapienza University in 2024, modifies YOLOv8 for faster object detection using the CLEVR-Hans dataset. We reduced layers and simplified components like C2f and SPPF to achieve up to 47% faster inference speeds and a model size 88% smaller, with only a minimal drop in accuracy.

The dataset contains objects like cubes, spheres, and cylinders, varying in size, color, and material, providing a challenging task for the model. Training was performed on Google Colab Pro+ using an A100 GPU, and all experiments, models, and outputs were stored in Google Drive.

For more details, you can read the full report here.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published