پروژه پایتون / ( 71997 )

Homework Assignment: Comparative Analysis of Panoptic Segmentation Backbones Objective: The objective of this homework is to gain hands-on experience with panoptic segmentation and compare the performance of different backbones on two diverse datasets: 7-Scenes and Cambridge Landmarks. Tasks: Setup: Clone the repository from the provided link: Panoptic-Deeplab. Follow the installation instructions in the repository's README file to set up the environment and dependencies. Download the 7-Scenes dataset from Microsoft's website and the Cambridge Landmarks dataset from Papers with Code. Experimentation: Use the provided codebase to perform panoptic segmentation on both the 7-Scenes and Cambridge Landmarks datasets. Experiment with at least three different backbone architectures for panoptic segmentation. Ensure that you use pre-trained weights for these backbones. Analysis and Reporting: Prepare a detailed report summarizing your findings. Compare the performance of the different backbones on both datasets. Discuss any observed differences in performance and computational efficiency. Include visualizations such as segmentation masks and qualitative analysis to support your conclusions. Provide insights into which backbone(s) perform better for each dataset and why. Discuss any challenges encountered during experimentation and how you addressed them. Submission Guidelines: Submit your report in PDF format. Include any necessary code snippets, configuration files, or additional materials used during experimentation. Ensure that your report is well-organized, clearly written, and properly formatted. Submit your assignment by the end of the semester. Note: To enhance your understanding, you are encouraged to explore additional resources and literature related to panoptic segmentation and backbone architectures. In your final presentation, you will be responsible for knowing what panoptic segmentation is and how it works in detail. رشته تحصیلی مهندسی کامپیوتر زمان یک ماه اینجا همه چی ذکر شده پایتون

Niko

امتیاز : 0/5

لیست پیشنهادها